fix rewards points (#10914)

* fix rewards points

* fixups

* * verify that we don't spend more in rewards than we've allocated for rewards
* purge f64s from calculations that could be done with integers

* test typical values

* simplify iteration over delegations some

* fixups

* Use try_from

* Add a comment for commission_split()

* Add assertion to detect inconsistent reward dist.

* Fix vote_balance_and_staked

* Don't overwrite accounts with stale copies

* Fix CI...

* Add tests for vote_balance_and_staked

* Add test for the determinism of update_rewards

* Revert "Don't overwrite accounts with stale copies"

This reverts commit 9886d085a69d695f48970f48f2f567e62ba05414.

* Make stake_delegation_accounts to return hashmap

Co-authored-by: Ryo Onodera <ryoqun@gmail.com>

programs/stake/src/stake_state.rs

@@ -19,7 +19,7 @@ use solana_sdk::{
     stake_history::{StakeHistory, StakeHistoryEntry},
 };
 use solana_vote_program::vote_state::{VoteState, VoteStateVersions};
-use std::collections::HashSet;
+use std::{collections::HashSet, convert::TryFrom};
 
 #[derive(Debug, Serialize, Deserialize, PartialEq, Clone, Copy, AbiExample)]
 #[allow(clippy::large_enum_variant)]
@@ -362,6 +362,15 @@ impl Authorized {
     }
 }
 
+/// captures a rewards round as lamports to be awarded
+///  and the total points over which those lamports
+///  are to be distributed
+//  basically read as rewards/points, but in integers instead of as an f64
+pub struct PointValue {
+    pub rewards: u64, // lamports to split
+    pub points: u128, // over these points
+}
+
 impl Stake {
     pub fn stake(&self, epoch: Epoch, history: Option<&StakeHistory>) -> u64 {
         self.delegation.stake(epoch, history)
@@ -369,36 +378,41 @@ impl Stake {
 
     pub fn redeem_rewards(
         &mut self,
-        point_value: f64,
+        point_value: &PointValue,
         vote_state: &VoteState,
         stake_history: Option<&StakeHistory>,
     ) -> Option<(u64, u64)> {
         self.calculate_rewards(point_value, vote_state, stake_history)
-            .map(|(voters_reward, stakers_reward, credits_observed)| {
+            .map(|(stakers_reward, voters_reward, credits_observed)| {
                 self.credits_observed = credits_observed;
                 self.delegation.stake += stakers_reward;
-                (voters_reward, stakers_reward)
+                (stakers_reward, voters_reward)
             })
     }
 
-    /// for a given stake and vote_state, calculate what distributions and what updates should be made
+    pub fn calculate_points(
-    /// returns a tuple in the case of a payout of:
-    ///   * voter_rewards to be distributed
-    ///   * staker_rewards to be distributed
-    ///   * new value for credits_observed in the stake
-    //  returns None if there's no payout or if any deserved payout is < 1 lamport
-    pub fn calculate_rewards(
         &self,
-        point_value: f64,
         vote_state: &VoteState,
         stake_history: Option<&StakeHistory>,
-    ) -> Option<(u64, u64, u64)> {
+    ) -> u128 {
+        self.calculate_points_and_credits(vote_state, stake_history)
+            .0
+    }
+
+    /// for a given stake and vote_state, calculate how many
+    ///   points were earned (credits * stake) and new value
+    ///   for credits_observed were the points paid
+    pub fn calculate_points_and_credits(
+        &self,
+        vote_state: &VoteState,
+        stake_history: Option<&StakeHistory>,
+    ) -> (u128, u64) {
         if self.credits_observed >= vote_state.credits() {
-            return None;
+            return (0, 0);
         }
 
         let mut credits_observed = self.credits_observed;
-        let mut total_rewards = 0f64;
+        let mut points = 0u128;
         for (epoch, credits, prev_credits) in vote_state.epoch_credits() {
             // figure out how much this stake has seen that
             //   for which the vote account has a record
@@ -414,29 +428,56 @@ impl Stake {
                 0
             };
 
-            total_rewards +=
+            points += u128::from(self.delegation.stake(*epoch, stake_history))
-                (self.delegation.stake(*epoch, stake_history) * epoch_credits) as f64 * point_value;
+                * u128::from(epoch_credits);
 
             // don't want to assume anything about order of the iterator...
             credits_observed = credits_observed.max(*credits);
         }
-        // don't bother trying to collect fractional lamports
+        (points, credits_observed)
-        if total_rewards < 1f64 {
+    }
+
+    /// for a given stake and vote_state, calculate what distributions and what updates should be made
+    /// returns a tuple in the case of a payout of:
+    ///   * staker_rewards to be distributed
+    ///   * voter_rewards to be distributed
+    ///   * new value for credits_observed in the stake
+    //  returns None if there's no payout or if any deserved payout is < 1 lamport
+    pub fn calculate_rewards(
+        &self,
+        point_value: &PointValue,
+        vote_state: &VoteState,
+        stake_history: Option<&StakeHistory>,
+    ) -> Option<(u64, u64, u64)> {
+        let (points, credits_observed) =
+            self.calculate_points_and_credits(vote_state, stake_history);
+
+        if points == 0 || point_value.points == 0 {
             return None;
         }
 
-        let (voter_rewards, staker_rewards, is_split) = vote_state.commission_split(total_rewards);
+        let rewards = points
+            .checked_mul(u128::from(point_value.rewards))
+            .unwrap()
+            .checked_div(point_value.points)
+            .unwrap();
 
-        if (voter_rewards < 1f64 || staker_rewards < 1f64) && is_split {
+        let rewards = u64::try_from(rewards).unwrap();
-            // don't bother trying to collect fractional lamports
+
+        // don't bother trying to split if fractional lamports got truncated
+        if rewards == 0 {
+            return None;
+        }
+        let (voter_rewards, staker_rewards, is_split) = vote_state.commission_split(rewards);
+
+        if (voter_rewards == 0 || staker_rewards == 0) && is_split {
+            // don't collect if we lose a whole lamport somewhere
+            //  is_split means there should be tokens on both sides,
+            //  uncool to move credits_observed if one side didn't get paid
             return None;
         }
 
-        Some((
+        Some((staker_rewards, voter_rewards, credits_observed))
-            voter_rewards as u64,
-            staker_rewards as u64,
-            credits_observed,
-        ))
     }
 
     fn redelegate(
@@ -836,17 +877,18 @@ impl<'a> StakeAccount for KeyedAccount<'a> {
 }
 
 // utility function, used by runtime
+// returns a tuple of (stakers_reward,voters_reward)
 pub fn redeem_rewards(
     stake_account: &mut Account,
     vote_account: &mut Account,
-    point_value: f64,
+    point_value: &PointValue,
     stake_history: Option<&StakeHistory>,
 ) -> Result<(u64, u64), InstructionError> {
     if let StakeState::Stake(meta, mut stake) = stake_account.state()? {
         let vote_state: VoteState =
             StateMut::<VoteStateVersions>::state(vote_account)?.convert_to_current();
 
-        if let Some((voters_reward, stakers_reward)) =
+        if let Some((stakers_reward, voters_reward)) =
             stake.redeem_rewards(point_value, &vote_state, stake_history)
         {
             stake_account.lamports += stakers_reward;
@@ -863,6 +905,22 @@ pub fn redeem_rewards(
     }
 }
 
+// utility function, used by runtime
+pub fn calculate_points(
+    stake_account: &Account,
+    vote_account: &Account,
+    stake_history: Option<&StakeHistory>,
+) -> Result<u128, InstructionError> {
+    if let StakeState::Stake(_meta, stake) = stake_account.state()? {
+        let vote_state: VoteState =
+            StateMut::<VoteStateVersions>::state(vote_account)?.convert_to_current();
+
+        Ok(stake.calculate_points(&vote_state, stake_history))
+    } else {
+        Err(InstructionError::InvalidAccountData)
+    }
+}
+
 // utility function, used by runtime::Stakes, tests
 pub fn new_stake_history_entry<'a, I>(
     epoch: Epoch,
@@ -923,6 +981,43 @@ pub fn create_account(
     vote_account: &Account,
     rent: &Rent,
     lamports: u64,
+) -> Account {
+    do_create_account(
+        authorized,
+        voter_pubkey,
+        vote_account,
+        rent,
+        lamports,
+        Epoch::MAX,
+    )
+}
+
+// utility function, used by tests
+pub fn create_account_with_activation_epoch(
+    authorized: &Pubkey,
+    voter_pubkey: &Pubkey,
+    vote_account: &Account,
+    rent: &Rent,
+    lamports: u64,
+    activation_epoch: Epoch,
+) -> Account {
+    do_create_account(
+        authorized,
+        voter_pubkey,
+        vote_account,
+        rent,
+        lamports,
+        activation_epoch,
+    )
+}
+
+fn do_create_account(
+    authorized: &Pubkey,
+    voter_pubkey: &Pubkey,
+    vote_account: &Account,
+    rent: &Rent,
+    lamports: u64,
+    activation_epoch: Epoch,
 ) -> Account {
     let mut stake_account = Account::new(lamports, std::mem::size_of::<StakeState>(), &id());
 
@@ -941,7 +1036,7 @@ pub fn create_account(
                 lamports - rent_exempt_reserve, // underflow is an error, is basically: assert!(lamports > rent_exempt_reserve);
                 voter_pubkey,
                 &vote_state,
-                std::u64::MAX,
+                activation_epoch,
                 &Config::default(),
             ),
         ))
@@ -954,7 +1049,7 @@ pub fn create_account(
 mod tests {
     use super::*;
     use crate::id;
-    use solana_sdk::{account::Account, pubkey::Pubkey, system_program};
+    use solana_sdk::{account::Account, native_token, pubkey::Pubkey, system_program};
     use solana_vote_program::vote_state;
     use std::cell::RefCell;
 
@@ -2171,7 +2266,14 @@ mod tests {
         // this one can't collect now, credits_observed == vote_state.credits()
         assert_eq!(
             None,
-            stake.redeem_rewards(1_000_000_000.0, &vote_state, None)
+            stake.redeem_rewards(
+                &PointValue {
+                    rewards: 1_000_000_000,
+                    points: 1
+                },
+                &vote_state,
+                None
+            )
         );
 
         // put 2 credits in at epoch 0
@@ -2180,8 +2282,15 @@ mod tests {
 
         // this one should be able to collect exactly 2
         assert_eq!(
-            Some((0, stake_lamports * 2)),
+            Some((stake_lamports * 2, 0)),
-            stake.redeem_rewards(1.0, &vote_state, None)
+            stake.redeem_rewards(
+                &PointValue {
+                    rewards: 1,
+                    points: 1
+                },
+                &vote_state,
+                None
+            )
         );
 
         assert_eq!(
@@ -2191,6 +2300,47 @@ mod tests {
         assert_eq!(stake.credits_observed, 2);
     }
 
+    #[test]
+    fn test_stake_state_calculate_points_with_typical_values() {
+        let mut vote_state = VoteState::default();
+
+        // bootstrap means fully-vested stake at epoch 0 with
+        //  10_000_000 SOL is a big but not unreasaonable stake
+        let stake = Stake::new(
+            native_token::sol_to_lamports(10_000_000f64),
+            &Pubkey::default(),
+            &vote_state,
+            std::u64::MAX,
+            &Config::default(),
+        );
+
+        // this one can't collect now, credits_observed == vote_state.credits()
+        assert_eq!(
+            None,
+            stake.calculate_rewards(
+                &PointValue {
+                    rewards: 1_000_000_000,
+                    points: 1
+                },
+                &vote_state,
+                None
+            )
+        );
+
+        let epoch_slots: u128 = 14 * 24 * 3600 * 160;
+        // put 193,536,000 credits in at epoch 0, typical for a 14-day epoch
+        //  this loop takes a few seconds...
+        for _ in 0..epoch_slots {
+            vote_state.increment_credits(0);
+        }
+
+        // no overflow on points
+        assert_eq!(
+            u128::from(stake.delegation.stake) * epoch_slots,
+            stake.calculate_points(&vote_state, None)
+        );
+    }
+
     #[test]
     fn test_stake_state_calculate_rewards() {
         let mut vote_state = VoteState::default();
@@ -2207,7 +2357,14 @@ mod tests {
         // this one can't collect now, credits_observed == vote_state.credits()
         assert_eq!(
             None,
-            stake.calculate_rewards(1_000_000_000.0, &vote_state, None)
+            stake.calculate_rewards(
+                &PointValue {
+                    rewards: 1_000_000_000,
+                    points: 1
+                },
+                &vote_state,
+                None
+            )
         );
 
         // put 2 credits in at epoch 0
@@ -2216,15 +2373,29 @@ mod tests {
 
         // this one should be able to collect exactly 2
         assert_eq!(
-            Some((0, stake.delegation.stake * 2, 2)),
+            Some((stake.delegation.stake * 2, 0, 2)),
-            stake.calculate_rewards(1.0, &vote_state, None)
+            stake.calculate_rewards(
+                &PointValue {
+                    rewards: 2,
+                    points: 2 // all his
+                },
+                &vote_state,
+                None
+            )
         );
 
         stake.credits_observed = 1;
         // this one should be able to collect exactly 1 (already observed one)
         assert_eq!(
-            Some((0, stake.delegation.stake, 2)),
+            Some((stake.delegation.stake, 0, 2)),
-            stake.calculate_rewards(1.0, &vote_state, None)
+            stake.calculate_rewards(
+                &PointValue {
+                    rewards: 1,
+                    points: 1
+                },
+                &vote_state,
+                None
+            )
         );
 
         // put 1 credit in epoch 1
@@ -2233,16 +2404,30 @@ mod tests {
         stake.credits_observed = 2;
         // this one should be able to collect the one just added
         assert_eq!(
-            Some((0, stake.delegation.stake, 3)),
+            Some((stake.delegation.stake, 0, 3)),
-            stake.calculate_rewards(1.0, &vote_state, None)
+            stake.calculate_rewards(
+                &PointValue {
+                    rewards: 2,
+                    points: 2
+                },
+                &vote_state,
+                None
+            )
         );
 
         // put 1 credit in epoch 2
         vote_state.increment_credits(2);
         // this one should be able to collect 2 now
         assert_eq!(
-            Some((0, stake.delegation.stake * 2, 4)),
+            Some((stake.delegation.stake * 2, 0, 4)),
-            stake.calculate_rewards(1.0, &vote_state, None)
+            stake.calculate_rewards(
+                &PointValue {
+                    rewards: 2,
+                    points: 2
+                },
+                &vote_state,
+                None
+            )
         );
 
         stake.credits_observed = 0;
@@ -2250,13 +2435,20 @@ mod tests {
         // (2 credits at stake of 1) + (1 credit at a stake of 2)
         assert_eq!(
             Some((
-                0,
                 stake.delegation.stake * 2 // epoch 0
                     + stake.delegation.stake // epoch 1
                     + stake.delegation.stake, // epoch 2
+                0,
                 4
             )),
-            stake.calculate_rewards(1.0, &vote_state, None)
+            stake.calculate_rewards(
+                &PointValue {
+                    rewards: 4,
+                    points: 4
+                },
+                &vote_state,
+                None
+            )
         );
 
         // same as above, but is a really small commission out of 32 bits,
@@ -2264,12 +2456,26 @@ mod tests {
         vote_state.commission = 1;
         assert_eq!(
             None, // would be Some((0, 2 * 1 + 1 * 2, 4)),
-            stake.calculate_rewards(1.0, &vote_state, None)
+            stake.calculate_rewards(
+                &PointValue {
+                    rewards: 4,
+                    points: 4
+                },
+                &vote_state,
+                None
+            )
         );
         vote_state.commission = 99;
         assert_eq!(
             None, // would be Some((0, 2 * 1 + 1 * 2, 4)),
-            stake.calculate_rewards(1.0, &vote_state, None)
+            stake.calculate_rewards(
+                &PointValue {
+                    rewards: 4,
+                    points: 4
+                },
+                &vote_state,
+                None
+            )
         );
     }
 

programs/vote/src/vote_state/mod.rs

@@ -253,13 +253,21 @@ impl VoteState {
     ///
     ///  if commission calculation is 100% one way or other,
     ///   indicate with false for was_split
-    pub fn commission_split(&self, on: f64) -> (f64, f64, bool) {
+    pub fn commission_split(&self, on: u64) -> (u64, u64, bool) {
         match self.commission.min(100) {
-            0 => (0.0, on, false),
+            0 => (0, on, false),
-            100 => (on, 0.0, false),
+            100 => (on, 0, false),
             split => {
-                let mine = on * f64::from(split) / f64::from(100);
+                let on = u128::from(on);
-                (mine, on - mine, true)
+                // Calculate mine and theirs independently and symmetrically instead of
+                // using the remainder of the other to treat them strictly equally.
+                // This is also to cancel the rewarding if either of the parties
+                // should receive only fractional lamports, resulting in not being rewarded at all.
+                // Thus, note that we intentionally discard any residual fractional lamports.
+                let mine = on * u128::from(split) / 100u128;
+                let theirs = on * u128::from(100 - split) / 100u128;
+
+                (mine as u64, theirs as u64, true)
             }
         }
     }
@@ -1562,19 +1570,22 @@ mod tests {
     fn test_vote_state_commission_split() {
         let vote_state = VoteState::default();
 
-        assert_eq!(vote_state.commission_split(1.0), (0.0, 1.0, false));
+        assert_eq!(vote_state.commission_split(1), (0, 1, false));
 
         let mut vote_state = VoteState::default();
         vote_state.commission = std::u8::MAX;
-        assert_eq!(vote_state.commission_split(1.0), (1.0, 0.0, false));
+        assert_eq!(vote_state.commission_split(1), (1, 0, false));
+
+        vote_state.commission = 99;
+        assert_eq!(vote_state.commission_split(10), (9, 0, true));
+
+        vote_state.commission = 1;
+        assert_eq!(vote_state.commission_split(10), (0, 9, true));
 
         vote_state.commission = 50;
-        let (voter_portion, staker_portion, was_split) = vote_state.commission_split(10.0);
+        let (voter_portion, staker_portion, was_split) = vote_state.commission_split(10);
 
-        assert_eq!(
+        assert_eq!((voter_portion, staker_portion, was_split), (5, 5, true));
-            (voter_portion.round(), staker_portion.round(), was_split),
-            (5.0, 5.0, true)
-        );
     }
 
     #[test]

runtime/src/bank.rs

@@ -55,11 +55,12 @@ use solana_sdk::{
     timing::years_as_slots,
     transaction::{Result, Transaction, TransactionError},
 };
-use solana_stake_program::stake_state::{self, Delegation};
+use solana_stake_program::stake_state::{self, Delegation, PointValue};
 use solana_vote_program::{vote_instruction::VoteInstruction, vote_state::VoteState};
 use std::{
     cell::RefCell,
     collections::{HashMap, HashSet},
+    convert::TryFrom,
     mem,
     ops::RangeInclusive,
     path::PathBuf,
@@ -898,11 +899,13 @@ impl Bank {
             let inflation = self.inflation.read().unwrap();
 
             (*inflation).validator(year) * self.capitalization() as f64 * period
-        };
+        } as u64;
 
-        let validator_points = self.stakes.write().unwrap().claim_points();
+        let vote_balance_and_staked = self.stakes.read().unwrap().vote_balance_and_staked();
-        let validator_point_value =
+
-            self.check_point_value(validator_rewards / validator_points as f64);
+        let validator_point_value = self.pay_validator_rewards(validator_rewards);
+
+        // this sysvar could be retired...
         self.update_sysvar_account(&sysvar::rewards::id(), |account| {
             sysvar::rewards::create_account(
                 self.inherit_sysvar_account_balance(account),
@@ -910,65 +913,117 @@ impl Bank {
             )
         });
 
-        let validator_rewards = self.pay_validator_rewards(validator_point_value);
+        let validator_rewards_paid =
+            self.stakes.read().unwrap().vote_balance_and_staked() - vote_balance_and_staked;
+        if let Some(rewards) = self.rewards.as_ref() {
+            assert_eq!(
+                validator_rewards_paid,
+                u64::try_from(rewards.iter().map(|(_pubkey, reward)| reward).sum::<i64>()).unwrap()
+            );
+        }
+
+        // verify that we didn't pay any more than we expected to
+        assert!(validator_rewards >= validator_rewards_paid);
 
         self.capitalization
-            .fetch_add(validator_rewards as u64, Ordering::Relaxed);
+            .fetch_add(validator_rewards_paid, Ordering::Relaxed);
     }
 
-    /// iterate over all stakes, redeem vote credits for each stake we can
+    /// map stake delegations into resolved (pubkey, account) pairs
-    ///   successfully load and parse, return total payout
+    ///  returns a map (has to be copied) of loaded
-    fn pay_validator_rewards(&mut self, point_value: f64) -> u64 {
+    ///   ( Vec<(staker info)> (voter account) ) keyed by voter pubkey
-        let stake_history = self.stakes.read().unwrap().history().clone();
+    ///
-        let mut validator_rewards = HashMap::new();
+    fn stake_delegation_accounts(&self) -> HashMap<Pubkey, (Vec<(Pubkey, Account)>, Account)> {
+        let mut accounts = HashMap::new();
 
-        let total_validator_rewards = self
+        self.stakes
+            .read()
+            .unwrap()
             .stake_delegations()
             .iter()
-            .map(|(stake_pubkey, delegation)| {
+            .for_each(|(stake_pubkey, delegation)| {
                 match (
                     self.get_account(&stake_pubkey),
                     self.get_account(&delegation.voter_pubkey),
                 ) {
-                    (Some(mut stake_account), Some(mut vote_account)) => {
+                    (Some(stake_account), Some(vote_account)) => {
-                        let rewards = stake_state::redeem_rewards(
+                        let entry = accounts
-                            &mut stake_account,
+                            .entry(delegation.voter_pubkey)
-                            &mut vote_account,
+                            .or_insert((Vec::new(), vote_account));
-                            point_value,
+                        entry.0.push((*stake_pubkey, stake_account));
-                            Some(&stake_history),
-                        );
-                        if let Ok((stakers_reward, voters_reward)) = rewards {
-                            self.store_account(&stake_pubkey, &stake_account);
-                            self.store_account(&delegation.voter_pubkey, &vote_account);
-
-                            if voters_reward > 0 {
-                                *validator_rewards
-                                    .entry(delegation.voter_pubkey)
-                                    .or_insert(0i64) += voters_reward as i64;
-                            }
-
-                            if stakers_reward > 0 {
-                                *validator_rewards.entry(*stake_pubkey).or_insert(0i64) +=
-                                    stakers_reward as i64;
-                            }
-
-                            stakers_reward + voters_reward
-                        } else {
-                            debug!(
-                                "stake_state::redeem_rewards() failed for {}: {:?}",
-                                stake_pubkey, rewards
-                            );
-                            0
-                        }
                     }
-                    (_, _) => 0,
+                    (_, _) => {}
                 }
+            });
+
+        accounts
+    }
+
+    /// iterate over all stakes, redeem vote credits for each stake we can
+    ///   successfully load and parse, return total payout
+    fn pay_validator_rewards(&mut self, rewards: u64) -> f64 {
+        let stake_history = self.stakes.read().unwrap().history().clone();
+
+        let mut stake_delegation_accounts = self.stake_delegation_accounts();
+
+        let points: u128 = stake_delegation_accounts
+            .iter()
+            .flat_map(|(_vote_pubkey, (stake_group, vote_account))| {
+                stake_group
+                    .iter()
+                    .map(move |(_stake_pubkey, stake_account)| (stake_account, vote_account))
+            })
+            .map(|(stake_account, vote_account)| {
+                stake_state::calculate_points(&stake_account, &vote_account, Some(&stake_history))
+                    .unwrap_or(0)
             })
             .sum();
 
+        if points == 0 {
+            return 0.0;
+        }
+
+        let point_value = PointValue { rewards, points };
+
+        let mut rewards = HashMap::new();
+
+        // pay according to point value
+        for (vote_pubkey, (stake_group, vote_account)) in stake_delegation_accounts.iter_mut() {
+            let mut vote_account_changed = false;
+            for (stake_pubkey, stake_account) in stake_group.iter_mut() {
+                let redeemed = stake_state::redeem_rewards(
+                    stake_account,
+                    vote_account,
+                    &point_value,
+                    Some(&stake_history),
+                );
+                if let Ok((stakers_reward, voters_reward)) = redeemed {
+                    self.store_account(&stake_pubkey, &stake_account);
+                    vote_account_changed = true;
+
+                    if voters_reward > 0 {
+                        *rewards.entry(*vote_pubkey).or_insert(0i64) += voters_reward as i64;
+                    }
+
+                    if stakers_reward > 0 {
+                        *rewards.entry(*stake_pubkey).or_insert(0i64) += stakers_reward as i64;
+                    }
+                } else {
+                    debug!(
+                        "stake_state::redeem_rewards() failed for {}: {:?}",
+                        stake_pubkey, redeemed
+                    );
+                }
+            }
+
+            if vote_account_changed {
+                self.store_account(&vote_pubkey, &vote_account);
+            }
+        }
+
         assert_eq!(self.rewards, None);
-        self.rewards = Some(validator_rewards.drain().collect());
+        self.rewards = Some(rewards.drain().collect());
-        total_validator_rewards
+        point_value.rewards as f64 / point_value.points as f64
     }
 
     fn update_recent_blockhashes_locked(&self, locked_blockhash_queue: &BlockhashQueue) {
@@ -986,21 +1041,6 @@ impl Bank {
         self.update_recent_blockhashes_locked(&blockhash_queue);
     }
 
-    // If the point values are not `normal`, bring them back into range and
-    // set them to the last value or 0.
-    fn check_point_value(&self, mut validator_point_value: f64) -> f64 {
-        let rewards = sysvar::rewards::Rewards::from_account(
-            &self
-                .get_account(&sysvar::rewards::id())
-                .unwrap_or_else(|| sysvar::rewards::create_account(1, 0.0)),
-        )
-        .unwrap_or_else(Default::default);
-        if !validator_point_value.is_normal() {
-            validator_point_value = rewards.validator_point_value;
-        }
-        validator_point_value
-    }
-
     fn collect_fees(&self) {
         let collector_fees = self.collector_fees.load(Ordering::Relaxed) as u64;
 
@@ -4730,7 +4770,18 @@ mod tests {
         }
         bank.store_account(&vote_id, &vote_account);
 
-        let validator_points = bank.stakes.read().unwrap().points();
+        let validator_points: u128 = bank
+            .stake_delegation_accounts()
+            .iter()
+            .flat_map(|(_vote_pubkey, (stake_group, vote_account))| {
+                stake_group
+                    .iter()
+                    .map(move |(_stake_pubkey, stake_account)| (stake_account, vote_account))
+            })
+            .map(|(stake_account, vote_account)| {
+                stake_state::calculate_points(&stake_account, &vote_account, None).unwrap_or(0)
+            })
+            .sum();
 
         // put a child bank in epoch 1, which calls update_rewards()...
         let bank1 = Bank::new_from_parent(
@@ -4774,6 +4825,89 @@ mod tests {
         );
     }
 
+    fn do_test_bank_update_rewards_determinism() -> u64 {
+        // create a bank that ticks really slowly...
+        let bank = Arc::new(Bank::new(&GenesisConfig {
+            accounts: (0..42)
+                .map(|_| {
+                    (
+                        Pubkey::new_rand(),
+                        Account::new(1_000_000_000, 0, &Pubkey::default()),
+                    )
+                })
+                .collect(),
+            // set it up so the first epoch is a full year long
+            poh_config: PohConfig {
+                target_tick_duration: Duration::from_secs(
+                    SECONDS_PER_YEAR as u64
+                        / MINIMUM_SLOTS_PER_EPOCH as u64
+                        / DEFAULT_TICKS_PER_SLOT,
+                ),
+                hashes_per_tick: None,
+                target_tick_count: None,
+            },
+
+            ..GenesisConfig::default()
+        }));
+
+        // enable lazy rent collection because this test depends on rent-due accounts
+        // not being eagerly-collected for exact rewards calculation
+        bank.lazy_rent_collection.store(true, Ordering::Relaxed);
+
+        assert_eq!(bank.capitalization(), 42 * 1_000_000_000);
+        assert_eq!(bank.rewards, None);
+
+        let vote_id = Pubkey::new_rand();
+        let mut vote_account = vote_state::create_account(&vote_id, &Pubkey::new_rand(), 50, 100);
+        let (stake_id1, stake_account1) = crate::stakes::tests::create_stake_account(123, &vote_id);
+        let (stake_id2, stake_account2) = crate::stakes::tests::create_stake_account(456, &vote_id);
+
+        // set up accounts
+        bank.store_account(&stake_id1, &stake_account1);
+        bank.store_account(&stake_id2, &stake_account2);
+
+        // generate some rewards
+        let mut vote_state = Some(VoteState::from(&vote_account).unwrap());
+        for i in 0..MAX_LOCKOUT_HISTORY + 42 {
+            if let Some(v) = vote_state.as_mut() {
+                v.process_slot_vote_unchecked(i as u64)
+            }
+            let versioned = VoteStateVersions::Current(Box::new(vote_state.take().unwrap()));
+            VoteState::to(&versioned, &mut vote_account).unwrap();
+            bank.store_account(&vote_id, &vote_account);
+            match versioned {
+                VoteStateVersions::Current(v) => {
+                    vote_state = Some(*v);
+                }
+                _ => panic!("Has to be of type Current"),
+            };
+        }
+        bank.store_account(&vote_id, &vote_account);
+
+        // put a child bank in epoch 1, which calls update_rewards()...
+        let bank1 = Bank::new_from_parent(
+            &bank,
+            &Pubkey::default(),
+            bank.get_slots_in_epoch(bank.epoch()) + 1,
+        );
+        // verify that there's inflation
+        assert_ne!(bank1.capitalization(), bank.capitalization());
+
+        bank1.capitalization()
+    }
+
+    #[test]
+    fn test_bank_update_rewards_determinism() {
+        // The same reward should be distributed given same credits
+        let expected_capitalization = do_test_bank_update_rewards_determinism();
+        // Repeat somewhat large number of iterations to expose possible different behavior
+        // depending on the randamly-seeded HashMap ordering
+        for _ in 0..30 {
+            let actual_capitalization = do_test_bank_update_rewards_determinism();
+            assert_eq!(actual_capitalization, expected_capitalization);
+        }
+    }
+
     // Test that purging 0 lamports accounts works.
     #[test]
     fn test_purge_empty_accounts() {
@@ -6364,23 +6498,6 @@ mod tests {
         assert!(bank.is_delta.load(Ordering::Relaxed));
     }
 
-    #[test]
-    #[allow(clippy::float_cmp)]
-    fn test_check_point_value() {
-        let (genesis_config, _) = create_genesis_config(500);
-        let bank = Arc::new(Bank::new(&genesis_config));
-
-        // check that point values are 0 if no previous value was known and current values are not normal
-        assert_eq!(bank.check_point_value(std::f64::INFINITY), 0.0);
-
-        bank.store_account(
-            &sysvar::rewards::id(),
-            &sysvar::rewards::create_account(1, 1.0),
-        );
-        // check that point values are the previous value if current values are not normal
-        assert_eq!(bank.check_point_value(std::f64::INFINITY), 1.0);
-    }
-
     #[test]
     fn test_bank_get_program_accounts() {
         let (genesis_config, mint_keypair) = create_genesis_config(500);

runtime/src/serde_snapshot/tests.rs

@@ -282,7 +282,7 @@ mod test_bank_serialize {
 
     // These some what long test harness is required to freeze the ABI of
     // Bank's serialization due to versioned nature
-    #[frozen_abi(digest = "6MnT4MzuLHe4Uq96YaF3JF2gL1RvprzQHCaV9TaWgYLe")]
+    #[frozen_abi(digest = "FaZaic5p7bvdsKDxGJmaPVyp12AbAmURyYoGiUdx1Ksu")]
     #[derive(Serialize, AbiExample)]
     pub struct BankAbiTestWrapperFuture {
         #[serde(serialize_with = "wrapper_future")]
@@ -305,7 +305,7 @@ mod test_bank_serialize {
         .serialize(s)
     }
 
-    #[frozen_abi(digest = "92KVEUQ8PwKe5DgZ6AyDQGAi9pvi7kfHdMBE4hcs5EQ4")]
+    #[frozen_abi(digest = "9g4bYykzsC86fULgu9iUh4kpvb1pxvAmipvyZPChLhws")]
     #[derive(Serialize, AbiExample)]
     pub struct BankAbiTestWrapperLegacy {
         #[serde(serialize_with = "wrapper_legacy")]

runtime/src/stakes.rs

@@ -15,9 +15,8 @@ pub struct Stakes {
     /// stake_delegations
     stake_delegations: HashMap<Pubkey, Delegation>,
 
-    /// unclaimed points.
+    /// unused
-    //  a point is a credit multiplied by the stake
+    unused: u64,
-    points: u64,
 
     /// current epoch, used to calculate current stake
     epoch: Epoch,
@@ -49,7 +48,7 @@ impl Stakes {
 
             Stakes {
                 stake_delegations: self.stake_delegations.clone(),
-                points: self.points,
+                unused: self.unused,
                 epoch,
                 vote_accounts: self
                     .vote_accounts
@@ -88,6 +87,18 @@ impl Stakes {
             .sum()
     }
 
+    pub fn vote_balance_and_staked(&self) -> u64 {
+        self.stake_delegations
+            .iter()
+            .map(|(_, stake_delegation)| stake_delegation.stake)
+            .sum::<u64>()
+            + self
+                .vote_accounts
+                .iter()
+                .map(|(_pubkey, (_staked, vote_account))| vote_account.lamports)
+                .sum::<u64>()
+    }
+
     pub fn is_stake(account: &Account) -> bool {
         solana_vote_program::check_id(&account.owner)
             || solana_stake_program::check_id(&account.owner)
@@ -106,15 +117,6 @@ impl Stakes {
                     |v| v.0,
                 );
 
-                // count any increase in points, can only go forward
-                let old_credits = old
-                    .and_then(|(_stake, old_account)| VoteState::credits_from(old_account))
-                    .unwrap_or(0);
-
-                let credits = VoteState::credits_from(account).unwrap_or(old_credits);
-
-                self.points += credits.saturating_sub(old_credits) * stake;
-
                 self.vote_accounts.insert(*pubkey, (stake, account.clone()));
             }
         } else if solana_stake_program::check_id(&account.owner) {
@@ -176,18 +178,6 @@ impl Stakes {
             .and_then(|(_k, (_stake, account))| VoteState::from(account))
             .map(|vote_state| vote_state.node_pubkey)
     }
-
-    /// currently unclaimed points
-    pub fn points(&self) -> u64 {
-        self.points
-    }
-
-    /// "claims" points, resets points to 0
-    pub fn claim_points(&mut self) -> u64 {
-        let points = self.points;
-        self.points = 0;
-        points
-    }
 }
 
 #[cfg(test)]
@@ -195,9 +185,7 @@ pub mod tests {
     use super::*;
     use solana_sdk::{pubkey::Pubkey, rent::Rent};
     use solana_stake_program::stake_state;
-    use solana_vote_program::vote_state::{
+    use solana_vote_program::vote_state::{self, VoteState};
-        self, VoteState, VoteStateVersions, MAX_LOCKOUT_HISTORY,
-    };
 
     //  set up some dummies for a staked node     ((     vote      )  (     stake     ))
     pub fn create_staked_node_accounts(stake: u64) -> ((Pubkey, Account), (Pubkey, Account)) {
@@ -224,6 +212,38 @@ pub mod tests {
         )
     }
 
+    pub fn create_warming_staked_node_accounts(
+        stake: u64,
+        epoch: Epoch,
+    ) -> ((Pubkey, Account), (Pubkey, Account)) {
+        let vote_pubkey = Pubkey::new_rand();
+        let vote_account = vote_state::create_account(&vote_pubkey, &Pubkey::new_rand(), 0, 1);
+        (
+            (vote_pubkey, vote_account),
+            create_warming_stake_account(stake, epoch, &vote_pubkey),
+        )
+    }
+
+    //   add stake to a vote_pubkey                               (   stake    )
+    pub fn create_warming_stake_account(
+        stake: u64,
+        epoch: Epoch,
+        vote_pubkey: &Pubkey,
+    ) -> (Pubkey, Account) {
+        let stake_pubkey = Pubkey::new_rand();
+        (
+            stake_pubkey,
+            stake_state::create_account_with_activation_epoch(
+                &stake_pubkey,
+                &vote_pubkey,
+                &vote_state::create_account(&vote_pubkey, &Pubkey::new_rand(), 0, 1),
+                &Rent::free(),
+                stake,
+                epoch,
+            ),
+        )
+    }
+
     #[test]
     fn test_stakes_basic() {
         for i in 0..4 {
@@ -302,76 +322,6 @@ pub mod tests {
         assert_eq!(stakes.highest_staked_node(), Some(vote11_node_pubkey))
     }
 
-    #[test]
-    fn test_stakes_points() {
-        let mut stakes = Stakes::default();
-        stakes.epoch = 4;
-
-        let stake = 42;
-        assert_eq!(stakes.points(), 0);
-        assert_eq!(stakes.claim_points(), 0);
-        assert_eq!(stakes.claim_points(), 0);
-
-        let ((vote_pubkey, mut vote_account), (stake_pubkey, stake_account)) =
-            create_staked_node_accounts(stake);
-
-        stakes.store(&vote_pubkey, &vote_account);
-        stakes.store(&stake_pubkey, &stake_account);
-
-        assert_eq!(stakes.points(), 0);
-        assert_eq!(stakes.claim_points(), 0);
-
-        let mut vote_state = Some(VoteState::from(&vote_account).unwrap());
-        for i in 0..MAX_LOCKOUT_HISTORY + 42 {
-            if let Some(v) = vote_state.as_mut() {
-                v.process_slot_vote_unchecked(i as u64)
-            }
-            let versioned = VoteStateVersions::Current(Box::new(vote_state.take().unwrap()));
-            VoteState::to(&versioned, &mut vote_account).unwrap();
-            match versioned {
-                VoteStateVersions::Current(v) => {
-                    vote_state = Some(*v);
-                }
-                _ => panic!("Has to be of type Current"),
-            };
-            stakes.store(&vote_pubkey, &vote_account);
-            assert_eq!(
-                stakes.points(),
-                vote_state.as_ref().unwrap().credits() * stake
-            );
-        }
-        vote_account.lamports = 0;
-        stakes.store(&vote_pubkey, &vote_account);
-        assert_eq!(
-            stakes.points(),
-            vote_state.as_ref().unwrap().credits() * stake
-        );
-
-        assert_eq!(
-            stakes.claim_points(),
-            vote_state.as_ref().unwrap().credits() * stake
-        );
-        assert_eq!(stakes.claim_points(), 0);
-        assert_eq!(stakes.claim_points(), 0);
-
-        // points come out of nowhere, but don't care here ;)
-        vote_account.lamports = 1;
-        stakes.store(&vote_pubkey, &vote_account);
-        assert_eq!(
-            stakes.points(),
-            vote_state.as_ref().unwrap().credits() * stake
-        );
-
-        // test going backwards, should never go backwards
-        let old_vote_state = vote_state;
-        let vote_account = vote_state::create_account(&vote_pubkey, &Pubkey::new_rand(), 0, 1);
-        stakes.store(&vote_pubkey, &vote_account);
-        assert_eq!(
-            stakes.points(),
-            old_vote_state.as_ref().unwrap().credits() * stake
-        );
-    }
-
     #[test]
     fn test_stakes_vote_account_disappear_reappear() {
         let mut stakes = Stakes::default();
@@ -528,4 +478,43 @@ pub mod tests {
             assert_eq!(vote_accounts.get(&vote_pubkey).unwrap().0, 0);
         }
     }
+
+    #[test]
+    fn test_vote_balance_and_staked_empty() {
+        let stakes = Stakes::default();
+        assert_eq!(stakes.vote_balance_and_staked(), 0);
+    }
+
+    #[test]
+    fn test_vote_balance_and_staked_normal() {
+        let mut stakes = Stakes::default();
+        impl Stakes {
+            pub fn vote_balance_and_warmed_staked(&self) -> u64 {
+                self.vote_accounts
+                    .iter()
+                    .map(|(_pubkey, (staked, account))| staked + account.lamports)
+                    .sum()
+            }
+        }
+
+        let genesis_epoch = 0;
+        let ((vote_pubkey, vote_account), (stake_pubkey, stake_account)) =
+            create_warming_staked_node_accounts(10, genesis_epoch);
+        stakes.store(&vote_pubkey, &vote_account);
+        stakes.store(&stake_pubkey, &stake_account);
+
+        assert_eq!(stakes.vote_balance_and_staked(), 11);
+        assert_eq!(stakes.vote_balance_and_warmed_staked(), 1);
+
+        for (epoch, expected_warmed_stake) in ((genesis_epoch + 1)..=3).zip(&[2, 3, 4]) {
+            stakes = stakes.clone_with_epoch(epoch);
+            // vote_balance_and_staked() always remain to return same lamports
+            // while vote_balance_and_warmed_staked() gradually increases
+            assert_eq!(stakes.vote_balance_and_staked(), 11);
+            assert_eq!(
+                stakes.vote_balance_and_warmed_staked(),
+                *expected_warmed_stake
+            );
+        }
+    }
 }