TVM: Discrepancy in HASHEXT_KECCAK256: Ignored Cell Refs Cause Inconsistencies with EVM Keccak-256

[yongjun925]

Describe the issue

The HASHEXT_KECCAK256 implementation in TON differs from the Keccak-256 implementation in the EVM. In TON, the HASHEXT_KECCAK256 does not handle Cell refs, leading to a discrepancy in hash results compared to the EVM, which could potentially cause hash conflicts.

What behavior did you expect?

I expected HASHEXT_KECCAK256 in TON to process Cell refs similarly to the EVM’s Keccak-256 implementation, ensuring consistent hash results across both platforms.

What was the actual behavior?

The current behavior of HASHEXT_KECCAK256 in TON only hashes the Cell data without taking the refs into account, resulting in different hash values compared to the EVM’s Keccak-256 for the same data structure.

How reliably can you reproduce the issue, what are the steps to do so?

(int) keccack256_single(slice _t0) asm "1 PUSHINT HASHEXT_KECCAK256";
int op = in_msg_body~load_uint(32);
if (op == op::keccack256()) {
dump(in_msg_bodyload_uint(64));
~dump(in_msg_body);
~dump(keccack256_single(in_msg_body));
}

messageBody1 x{0000006800000000000000000300007DDE0000334B0000000000000000000000008157668EC72C279C20C9D7387B7B711FCF713A4D0000271AEFAB5D355AD91A98BCBACEA3854E0EECB83942801BE6722B9A941317F050E2A668656C6C6F20776F726C642068656C6C6F20776F726C64202168656C6C6F20776F726C642068}
x{656C6C6F20776F726C64202168656C6C6F20776F726C642068656C6C6F20776F726C642021}
Output result1: 112128826363682873612496808615290492480709047111283954187881755354176156987312

messageBody2 x{0000006800000000000000000300007DDE0000334B0000000000000000000000008157668EC72C279C20C9D7387B7B711FCF713A4D0000271AEFAB5D355AD91A98BCBACEA3854E0EECB83942801BE6722B9A941317F050E2A668656C6C6F20776F726C642068656C6C6F20776F726C64202168656C6C6F20776F726C642068}
x{656C6C6F20776F726C64202168656C6C6F20776F726C642068656C6C6F20776F726C64202168656C6C6F20776F726C642068656C6C6F20776F726C64202168656C6C6F20776F726C642068656C6C6F20776F726C642021}
Output result2: 112128826363682873612496808615290492480709047111283954187881755354176156987312

hexToBigInt(
keccak256(
'0x0300007DDE0000334B0000000000000000000000008157668EC72C279C20C9D7387B7B711FCF713A4D0000271AEFAB5D355AD91A98BCBACEA3854E0EECB83942801BE6722B9A941317F050E2A668656C6C6F20776F726C642068656C6C6F20776F726C64202168656C6C6F20776F726C642068' as '0x',
),
)
Output result3: 112128826363682873612496808615290492480709047111283954187881755354176156987312