Attackathon _ Fuel Network 32275 - [Smart Contract - Medium] Various Sway Libs Bugs

Submitted on Mon Jun 17 2024 08:57:19 GMT-0400 (Atlantic Standard Time) by @anatomist for Attackathon | Fuel Network

Report ID: #32275

Report type: Smart Contract

Report severity: Medium

Target: https://github.com/FuelLabs/sway-libs/tree/0f47d33d6e5da25f782fc117d4be15b7b12d291b

Impacts:

• Incorrect library function behaviors

Description

Brief/Intro

sway-libs provides a lot of utility functions to help contract development. Thus its correctness is important to ensure developers don't suffer from hidden vulnerabilities. We've identified a large amount of bugs within the libraries, and will go through each bug class in this report.

Vulnerability Details

Incorrect bit width for signed_integers and fixed_point

• The i256::bits is incorrectly set to 128.

• Strictly speaking, ifp64 has ~33 bits and not 64 bits. Similarly ifp128 has ~64 bits and ifp256 has ~129 bits

Incorrect two's complement implementation for signed_integers

• The twos complement currently doesn't make sense. For instance, two's complement for -1i8 should be 0xff, but the implementation returns 0x81

Incorrect indent for i128 and i256

• i128 indent should be 1<<127, and i256 indent should be 1<<255

Unnecessary subtract In UFP32::fract

• The subtraction of u32::max in UFP32::fract will lead to underflow and panic

Incorrect usage of IFP::from

• IFP::from treats input as raw underlying instead of integers (different from signed_integer). It is misused in several places

IFP::ceil overflow on values close to min

• IFP::ceil takes ceilof underlying and then adjusts the values. Thus values near tomin` might suffer from incorrect overflow

Incorrect panic of fixed_point::round

• fixed_point::round takes both ceil / floor and use them to derive the rounded values. Thus even if the final result falls within the valid range, function would still panic if either ceil of floor does.

Incorrect denominator in UFP128::sqrt and UFP32::sqrt

• The denominator for UFP128::sqrt should be 1<<32 instead of 2<<32

• The denominator for UFP32::sqrt should be 1<<8 instead of 1<<16

UFP::sqrt precision loss

• The implementation of UFP::sqrt loses 1/4 of precision in terms of its size.

UFP::pow` premature overflow leads to revert

• UFP::pow directly calls pow on the underlying value, which will easily overflow for values >= 1.

fixed_point::exp precision loss

• UFP::exp is wildly inaccurate. We haven't properly estimated to amount of precision lost. But the PoC attached should demonstrate the idea.

UFP32::exp and IFP256::exp uses incorrect taylor series

• UFP32::exp uses the taylor series for UFP64::exp

• IFP256::exp uses the taylor series for IFP128::exp

Incorrect min for IFP

• min for IFP should use type::MAX as underlying value

Lack of consideration of negative zero for IFP compare functions

• IFP::gt and IFP::lt doesn't consider negative zeros. This might lead to incorrect results.

• IFP::non_negative doesn't consider negative zeroes. And might return false for those.

Incorrect adjustment for IFP::ceil

• IFP::ceil increases underlying when the value is negative, while the correct implementation should be to decrease underlying.

Impact Details

It is hard to provide a concrete impact for library functions. But let's just say users may not be able to predict its behavior and might be caught of guard when those functions do unexpected stuff. And given the abundance of bugs, it seems extremely likely that users would run into those.

References

Incorrect bit width for signed_integers and fixed_point

• https://github.com/FuelLabs/sway-libs/blob/0f47d33d6e5da25f782fc117d4be15b7b12d291b/libs/src/signed_integers/i256.sw#L88

• https://github.com/FuelLabs/sway-libs/blob/0f47d33d6e5da25f782fc117d4be15b7b12d291b/libs/src/fixed_point/ifp64.sw#L46

• https://github.com/FuelLabs/sway-libs/blob/0f47d33d6e5da25f782fc117d4be15b7b12d291b/libs/src/fixed_point/ifp128.sw#L46

• https://github.com/FuelLabs/sway-libs/blob/0f47d33d6e5da25f782fc117d4be15b7b12d291b/libs/src/fixed_point/ifp256.sw#L46

Incorrect two's complement implementation for signed_integers

• https://github.com/FuelLabs/sway-libs/blob/0f47d33d6e5da25f782fc117d4be15b7b12d291b/libs/src/signed_integers/i8.sw#L398

• https://github.com/FuelLabs/sway-libs/blob/0f47d33d6e5da25f782fc117d4be15b7b12d291b/libs/src/signed_integers/i16.sw#L399

• https://github.com/FuelLabs/sway-libs/blob/0f47d33d6e5da25f782fc117d4be15b7b12d291b/libs/src/signed_integers/i32.sw#L398

• https://github.com/FuelLabs/sway-libs/blob/0f47d33d6e5da25f782fc117d4be15b7b12d291b/libs/src/signed_integers/i64.sw#L399

• https://github.com/FuelLabs/sway-libs/blob/0f47d33d6e5da25f782fc117d4be15b7b12d291b/libs/src/signed_integers/i128.sw#L423

• https://github.com/FuelLabs/sway-libs/blob/0f47d33d6e5da25f782fc117d4be15b7b12d291b/libs/src/signed_integers/i256.sw#L419

Incorrect indent for i128 and i256

• https://github.com/FuelLabs/sway-libs/blob/0f47d33d6e5da25f782fc117d4be15b7b12d291b/libs/src/signed_integers/i128.sw#L38

• https://github.com/FuelLabs/sway-libs/blob/0f47d33d6e5da25f782fc117d4be15b7b12d291b/libs/src/signed_integers/i256.sw#L39

Unnecessary subtract In UFP32::fract

• https://github.com/FuelLabs/sway-libs/blob/0f47d33d6e5da25f782fc117d4be15b7b12d291b/libs/src/fixed_point/ufp32.sw#L391

Incorrect usage of IFP::from

• https://github.com/FuelLabs/sway-libs/blob/0f47d33d6e5da25f782fc117d4be15b7b12d291b/libs/src/fixed_point/ifp64.sw#L416

• https://github.com/FuelLabs/sway-libs/blob/0f47d33d6e5da25f782fc117d4be15b7b12d291b/libs/src/fixed_point/ifp64.sw#L475

• https://github.com/FuelLabs/sway-libs/blob/0f47d33d6e5da25f782fc117d4be15b7b12d291b/libs/src/fixed_point/ifp64.sw#L476

• https://github.com/FuelLabs/sway-libs/blob/0f47d33d6e5da25f782fc117d4be15b7b12d291b/libs/src/fixed_point/ifp128.sw#L416

• https://github.com/FuelLabs/sway-libs/blob/0f47d33d6e5da25f782fc117d4be15b7b12d291b/libs/src/fixed_point/ifp128.sw#L475

• https://github.com/FuelLabs/sway-libs/blob/0f47d33d6e5da25f782fc117d4be15b7b12d291b/libs/src/fixed_point/ifp128.sw#L476

• https://github.com/FuelLabs/sway-libs/blob/0f47d33d6e5da25f782fc117d4be15b7b12d291b/libs/src/fixed_point/ifp256.sw#L554

IFP::ceil overflow on values close to min

• https://github.com/FuelLabs/sway-libs/blob/0f47d33d6e5da25f782fc117d4be15b7b12d291b/libs/src/fixed_point/ifp64.sw#L473

• https://github.com/FuelLabs/sway-libs/blob/0f47d33d6e5da25f782fc117d4be15b7b12d291b/libs/src/fixed_point/ifp128.sw#L473

• https://github.com/FuelLabs/sway-libs/blob/0f47d33d6e5da25f782fc117d4be15b7b12d291b/libs/src/fixed_point/ifp256.sw#L473

Incorrect panic of fixed_point::round

• https://github.com/FuelLabs/sway-libs/blob/0f47d33d6e5da25f782fc117d4be15b7b12d291b/libs/src/fixed_point/ufp32.sw#L443

• https://github.com/FuelLabs/sway-libs/blob/0f47d33d6e5da25f782fc117d4be15b7b12d291b/libs/src/fixed_point/ufp64.sw#L437

• https://github.com/FuelLabs/sway-libs/blob/0f47d33d6e5da25f782fc117d4be15b7b12d291b/libs/src/fixed_point/ufp128.sw#L383

• https://github.com/FuelLabs/sway-libs/blob/0f47d33d6e5da25f782fc117d4be15b7b12d291b/libs/src/fixed_point/ifp64.sw#L515

• https://github.com/FuelLabs/sway-libs/blob/0f47d33d6e5da25f782fc117d4be15b7b12d291b/libs/src/fixed_point/ifp128.sw#L515

• https://github.com/FuelLabs/sway-libs/blob/0f47d33d6e5da25f782fc117d4be15b7b12d291b/libs/src/fixed_point/ifp256.sw#L515

Incorrect denominator in UFP128::sqrt and UFP32::sqrt

• https://github.com/FuelLabs/sway-libs/blob/0f47d33d6e5da25f782fc117d4be15b7b12d291b/libs/src/fixed_point/ufp128.sw#L443

• https://github.com/FuelLabs/sway-libs/blob/0f47d33d6e5da25f782fc117d4be15b7b12d291b/libs/src/fixed_point/ufp32.sw#L462

UFP::sqrt precision loss

• https://github.com/FuelLabs/sway-libs/blob/0f47d33d6e5da25f782fc117d4be15b7b12d291b/libs/src/fixed_point/ufp32.sw#L462

• https://github.com/FuelLabs/sway-libs/blob/0f47d33d6e5da25f782fc117d4be15b7b12d291b/libs/src/fixed_point/ufp64.sw#L456

• https://github.com/FuelLabs/sway-libs/blob/0f47d33d6e5da25f782fc117d4be15b7b12d291b/libs/src/fixed_point/ufp128.sw#L442

UFP::pow` premature overflow leads to revert

• https://github.com/FuelLabs/sway-libs/blob/0f47d33d6e5da25f782fc117d4be15b7b12d291b/libs/src/fixed_point/ufp32.sw#L492

• https://github.com/FuelLabs/sway-libs/blob/0f47d33d6e5da25f782fc117d4be15b7b12d291b/libs/src/fixed_point/ufp32.sw#L497

• https://github.com/FuelLabs/sway-libs/blob/0f47d33d6e5da25f782fc117d4be15b7b12d291b/libs/src/fixed_point/ufp64.sw#L487

• https://github.com/FuelLabs/sway-libs/blob/0f47d33d6e5da25f782fc117d4be15b7b12d291b/libs/src/fixed_point/ufp128.sw#L450

• https://github.com/FuelLabs/sway-libs/blob/0f47d33d6e5da25f782fc117d4be15b7b12d291b/libs/src/fixed_point/ufp128.sw#L452

fixed_point::exp precision loss

• https://github.com/FuelLabs/sway-libs/blob/0f47d33d6e5da25f782fc117d4be15b7b12d291b/libs/src/fixed_point/ufp32.sw#L471

• https://github.com/FuelLabs/sway-libs/blob/0f47d33d6e5da25f782fc117d4be15b7b12d291b/libs/src/fixed_point/ufp64.sw#L465

• https://github.com/FuelLabs/sway-libs/blob/0f47d33d6e5da25f782fc117d4be15b7b12d291b/libs/src/fixed_point/ufp128.sw#L469

• https://github.com/FuelLabs/sway-libs/blob/0f47d33d6e5da25f782fc117d4be15b7b12d291b/libs/src/fixed_point/ifp64.sw#L533

• https://github.com/FuelLabs/sway-libs/blob/0f47d33d6e5da25f782fc117d4be15b7b12d291b/libs/src/fixed_point/ifp128.sw#L533

• https://github.com/FuelLabs/sway-libs/blob/0f47d33d6e5da25f782fc117d4be15b7b12d291b/libs/src/fixed_point/ifp256.sw#L533

UFP32::exp and IFP256::exp uses incorrect taylor series

• https://github.com/FuelLabs/sway-libs/blob/0f47d33d6e5da25f782fc117d4be15b7b12d291b/libs/src/fixed_point/ufp32.sw#L475

• https://github.com/FuelLabs/sway-libs/blob/0f47d33d6e5da25f782fc117d4be15b7b12d291b/libs/src/fixed_point/ifp256.sw#L537

Incorrect min for IFP

• https://github.com/FuelLabs/sway-libs/blob/0f47d33d6e5da25f782fc117d4be15b7b12d291b/libs/src/fixed_point/ifp64.sw#L87

• https://github.com/FuelLabs/sway-libs/blob/0f47d33d6e5da25f782fc117d4be15b7b12d291b/libs/src/fixed_point/ifp128.sw#L87

• https://github.com/FuelLabs/sway-libs/blob/0f47d33d6e5da25f782fc117d4be15b7b12d291b/libs/src/fixed_point/ifp256.sw#L87

Lack of consideration of negative zero for IFP compare functions

• https://github.com/FuelLabs/sway-libs/blob/0f47d33d6e5da25f782fc117d4be15b7b12d291b/libs/src/fixed_point/ifp64.sw#L208

• https://github.com/FuelLabs/sway-libs/blob/0f47d33d6e5da25f782fc117d4be15b7b12d291b/libs/src/fixed_point/ifp64.sw#L220

• https://github.com/FuelLabs/sway-libs/blob/0f47d33d6e5da25f782fc117d4be15b7b12d291b/libs/src/fixed_point/ifp128.sw#L208

• https://github.com/FuelLabs/sway-libs/blob/0f47d33d6e5da25f782fc117d4be15b7b12d291b/libs/src/fixed_point/ifp128.sw#L220

• https://github.com/FuelLabs/sway-libs/blob/0f47d33d6e5da25f782fc117d4be15b7b12d291b/libs/src/fixed_point/ifp256.sw#L208

• https://github.com/FuelLabs/sway-libs/blob/0f47d33d6e5da25f782fc117d4be15b7b12d291b/libs/src/fixed_point/ifp256.sw#L220

• https://github.com/FuelLabs/sway-libs/blob/0f47d33d6e5da25f782fc117d4be15b7b12d291b/libs/src/fixed_point/ifp64.sw#L193

• https://github.com/FuelLabs/sway-libs/blob/0f47d33d6e5da25f782fc117d4be15b7b12d291b/libs/src/fixed_point/ifp128.sw#L193

• https://github.com/FuelLabs/sway-libs/blob/0f47d33d6e5da25f782fc117d4be15b7b12d291b/libs/src/fixed_point/ifp256.sw#L193

Incorrect adjustment for IFP::ceil

• https://github.com/FuelLabs/sway-libs/blob/0f47d33d6e5da25f782fc117d4be15b7b12d291b/libs/src/fixed_point/ifp64.sw#L475

• https://github.com/FuelLabs/sway-libs/blob/0f47d33d6e5da25f782fc117d4be15b7b12d291b/libs/src/fixed_point/ifp128.sw#L475

• https://github.com/FuelLabs/sway-libs/blob/0f47d33d6e5da25f782fc117d4be15b7b12d291b/libs/src/fixed_point/ifp256.sw#L475

Proof of concept

Proof of Concept

Incorrect bit width for signed_integers and fixed_point

#[test]
fn i256_bits() -> () {
    assert(I256::bits() == 256);
    ()
}

Incorrect two's complement implementation for signed_integers

#[test]
fn i8_twos_compliment() -> () {
    let val = I8::neg_from(1);
    let complement = val.twos_complement();
    assert(complement.underlying() == 255);
}

Incorrect indent for i128 and i256

#[test]
fn i128_indent() -> () {
    // -2^65 should be in range
    let u128_val = U128::from((2, 0));
    let i128_val = I128::neg_from(u128_val);
    ()
}

Unnecessary subtract In UFP32::fract

#[test]
fn ufp32_fract_incorrect_underflow() -> () {
    let zero = UFP32::from_uint(0);
    let one = UFP32::from_uint(1);
    assert(one.fract() == zero);
    ()
}

Incorrect usage of IFP::from

#[test]
fn ifp64_incorrect_usage_of_from() -> () {
    let ifp64_zero = IFP64::zero();
    let ifp64_val = ifp64_zero - IFP64::from(UFP32::from(65537u32));                    //-0x1.0001

    assert((ifp64_zero - IFP64::from_uint(2u32)) == ifp64_val.floor());
    ()
}

IFP::ceil overflow on values close to min

#[test]
fn ifp128_ceil_incorrect_overflow() -> () {
    let ifp128_zero = IFP128::zero();
    let ifp128_val = ifp128_zero - IFP128::from(UFP64::from(18446744069414584321));     //-0xffffffff.00000001
    let expected = ifp128_zero - IFP128::from(UFP64::from(18446744069414584320));       //-0xffffffff.00000000

    assert(expected == ifp64_val.ceil());
    ()
}

Incorrect panic of fixed_point::round

#[test]
fn ufp64_round_incorrect_overflow() -> () {
    let val = UFP64::from(18446744069414584321);        //0xffffffff.00000001
    assert(UFP64::from(18446744069414584320) == val.round());
    ()
}

Incorrect denominator in UFP128::sqrt and UFP32::sqrt

#[test]
fn ufp128_sqrt_incorrect_denom() -> () {
    let val = UFP128::from((0, 1));                             //0x0000000000000000.0000000000000001
    assert(UFP128::from((0, 4294967296)) == val.sqrt());        //0x0000000000000000.0000000100000000
    ()
}

UFP::sqrt precision loss

#[test]
fn ufp64_sqrt_precision_loss() -> () {
    let val = UFP64::from(3);                           //0x00000000.00000002
    assert(UFP64::from(92681) == val.sqrt());           //0x00000000.00016a09
    ()
}

UFP::pow` premature overflow leads to revert

#[test]
fn ufp64_pow_premature_overflow() -> () {
    let val = UFP64::from_uint(1);                      //0x00000001.00000000
    assert(val == val.pow(2u32));
    ()
}

fixed_point::exp precision loss

#[test]
fn u64_imprecise_exponent() -> () {
    let val = UFP64::from_uint(1);
    let val = UFP64::exp(val);
    assert(val == UFP64::from(11674931554));            //0x2.b7e15162
    ()
}

UFP32::exp and IFP256::exp uses incorrect taylor series

Incorrect min for IFP

#[test]
fn ifp128_incorrect_min() -> () {
    let ifp128_zero = IFP128::zero();
    let ifp128_min = IFP128::min();
    let ifp128_max = IFP128::max();

    assert(ifp128_zero - ifp128_max == ifp128_min);
    ()
}

Lack of consideration of negative zero for IFP compare functions

#[test]
fn ifp128_negative_zero() -> () {
    let val = IFP128::from(UFP64::from(1));
    let zero = IFP128::zero();
    let neg_zero = (zero - val).fract();
    assert((zero == neg_zero) != (zero > neg_zero));
    assert(neg_zero.non_negative());
    ()
}

Incorrect adjustment for IFP::ceil

#[test]
fn ifp128_incorrect_ceil() -> () {
    let ifp128_zero = IFP128::zero();
    let ifp128_val = ifp128_zero - IFP128::from(UFP64::from(1));

    assert(ifp128_val.ceil() > ifp128_val);
    ()
}