hickuphh3 - Loss of rewards if `pointsPerShare` increment exceeds `type(int256).max`

[sherlock-admin]

hickuphh3

medium

Loss of rewards if pointsPerShare increment exceeds type(int256).max

Summary

Rewards would be unrecoverable if pointsPerShare is incremented to more than type(int256).max.

Vulnerability Detail

When rewards are distributed by calling distributeRewards(), the internal function _distributeRewards() may cause pointsPerShare to overflow to a point where its multiplication with shares cannot be safely casted to int256.

POC

Assume

• pointsPerShare = 0
• shares = getTotalShares() = 1 and amount = (type(int256).max) / POINTS_MULTIPLIER + 1

Then, pointsPerShare becomes

_amount * POINTS_MULTIPLIER / shares
= (type(int256).max) / type(uint128).max + 1) * type(uint128).max / 1
= type(int256).max) + type(uint128).max 

This would then cause cumulativeRewardsOf() <- withdrawableRewardsOf() <- _prepareCollect() to revert ie. claiming of rewards to revert because the safecasting to int256 would revert:

return ((pointsPerShare * getSharesOf(_account)).toInt256() + pointsCorrection[_account]).toUint256() / POINTS_MULTIPLIER;

causing distributed rewards to be unclaimable.

Impact

Rewards can never be claimed nor recovered.

Code Snippet

https://github.com/Merit-Circle/merit-liquidity-mining/blob/ce5feaae19126079d309ac8dd9a81372648437f1/contracts/base/AbstractRewards.sol#L74

https://github.com/Merit-Circle/merit-liquidity-mining/blob/ce5feaae19126079d309ac8dd9a81372648437f1/contracts/base/AbstractRewards.sol#L95-L97

Tool used

Manual Review

Recommendation

In _distributeRewards(), ensure that pointsPerShare * shares can be safely casted to toInt256, ie. does not exceed uint256(type(int256).max).

if (_amount > 0) {
   pointsPerShare = pointsPerShare + (_amount * POINTS_MULTIPLIER / shares);
+  if (pointsPerShare * shares  > uint256(type(int256).max)) revert PointsOverflow();
   emit RewardsDistributed(msg.sender, _amount);
}

P.S the unsafe casting of pointsPerShare in _correctPoints will be fine because pointsPerShare < pointsPerShare * shares <= uint256(type(int256).max)

[jack-the-pug]

This should be invalid or downgraded to low/info.

Because _amount won't possibly be as large as (type(int256).max) / POINTS_MULTIPLIER + 1, that's an extremely large number: about 3e38.

And the manipulation of pointsPerShare is an issue itself: https://github.com/sherlock-audit/2022-10-merit-circle-judging/issues/103

[federava]

This is part of the design and expected behaviour. We think that such huge numbers will not be seen in the foreseeable future.