circuit for invalid creation code

[DreamWuGit]

Description

circuit for invalid creation code error in create, markdown spec privacy-scaling-explorations/zkevm-specs#400

Issue Link

issue #1291

Type of change

• New feature (non-breaking change which adds functionality)

Rationale

Get first byte of code to store , check it is 0xef

This PR contains:

• buss mapping: at the return opcode in create, get the first byte .
• add circuit & test
• add tx deploy trace test .

[hero78119]

Hi @DreamWuGit , would you kindly resolved the conflict, since this pr #1342 is merged :) ? Thanks!
I think few error handling step should be able to handled by new merged restore context function.

[lispc]

yes @DreamWuGit can copy new codes from scroll fork i think. i have already merged these branches.

[ed255]

@DreamWuGit kindly reminder: could you update this PR to resolve conflicts? Thanks!

[ChihChengLiang]

CREATE PRs (#1419 #1425 #1430 ) are all merged. This PR is unblocked.
@DreamWuGit

[lispc]

@silathdiir will work on this

[ChihChengLiang]

@silathdiir is this ready for review?

[silathdiir]

Yes, thanks. I update this PR with latest main branch, and also the spec doc PR privacy-scaling-explorations/zkevm-specs#400.

Please help review when you have time. Thanks.

[hero78119]

Hi, one question regarding to soundness

In create.rs and return_revert, I didn't find the constraint where force the prover must fill is_success == false if code starting with 0xef. I am wondering what if a malicious prover fill is_success = true, can the create/return_revert circuit catch up this InValidCreationWord error or actually it can pass?
From what I understand the only place to avoid this soundness is create.rs must also have constrain which force to go this error execution step.

[silathdiir]

Hi, one question regarding to soundness

In create.rs and return_revert, I didn't find the constraint where force the prover must fill is_success == false if code starting with 0xef. I am wondering what if a malicious prover fill is_success = true, can the create/return_revert circuit catch up this InValidCreationWord error or actually it can pass? From what I understand the only place to avoid this soundness is create.rs must also have constrain which force to go this error execution step.

Yes, I see, I think you are right.
I will add is_first_byte_invalid and constrain as below to create.rs and begin_tx.rs (which may also call Create) in this PR:

let is_first_byte_invalid = IsEqualGadget::construct(cb, first_byte.expr(), 0xef.expr());

cb.condition(is_success.expr(), |cb| {
    cb.require_zero(
        "init_code_first_byte != 0xef if success",
        is_first_byte_invalid
    );
});

[silathdiir]

Hi @hero78119, sorry for late update.

After some investigation, it seems that in CREATE (or CREATE2) execution step, the contract Run result could not be retrieved (from memory) to check this ErrInvalidCode error. Please correct me if I was wrong, thanks.

So I fix to add a constraint which constrains the first byte of create init code must not be 0xef to return_revert in this commit 3138114. Please help review again when you have time. Thanks.

[hero78119]

Thanks for the quick fix!! @silathdiir 🙏

I think we also need to fix specs part, other than that all LTGM!

[hero78119]

👍 LGTM! 💯

[ChihChengLiang]

Hi @lispc,
Would you be available to review this PR?
We can assign another reviewer if you are busy.

[lispc]

i will have a look