iframe-proxy

sueun-dev · 2026-03-08T20:52:44Z

Summary

carry each transaction's gas_limit() into the BAL-backed parallel execution result set
enforce the same remaining block gas check used by the sequential LEVM path when building parallel receipts
add a regression test for a later transaction whose gas_limit exceeds the remaining block gas

Testing

cargo test -p ethrex-vm

Copilot

Pull request overview

Aligns the Amsterdam BAL-backed parallel LEVM execution pipeline with the sequential path by enforcing the “remaining block gas” constraint based on cumulative post-refund gas (EIP-7778), preventing blocks from being accepted when a later tx’s gas_limit() exceeds remaining block gas.

Changes:

Carry tx.gas_limit() through the parallel execution result tuple.
Add build_parallel_receipts(...) to sort results, run check_gas_limit(...) during receipt construction, and return (receipts, block_gas_used).
Add a regression test covering a later transaction whose gas_limit exceeds remaining block gas.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

greptile-apps · 2026-03-08T20:56:37Z

Greptile Summary

This PR fixes a missing block gas limit validation in the parallel LEVM execution pipeline (Amsterdam+ forks). Previously, execute_block_parallel built receipts inline without calling check_gas_limit, meaning a block whose transactions collectively exceeded the gas limit could pass undetected in the parallel path while being correctly rejected in the sequential path.

Changes:

Extracts receipt-building logic into a new build_parallel_receipts helper that mirrors the sequential path's check_gas_limit call (using cumulative_gas_used / POST-REFUND gas, consistent with EIP-7778).
Extends the parallel execution tuple from (usize, TxType, ExecutionReport) to (usize, TxType, u64, ExecutionReport) to carry each transaction's gas_limit() into the receipt-building phase.
Adds a regression test that arranges out-of-order results and verifies the second transaction is rejected when its gas_limit exceeds the remaining block gas.

The logic is correct and symmetric with the sequential path: check_gas_limit is invoked before updating cumulative_gas_used, and both cumulative_gas_used (gas_spent, POST-REFUND) and block_gas_used (gas_used, PRE-REFUND) are tracked independently per EIP-7778.

Confidence Score: 5/5

This PR is safe to merge — it is a targeted, well-tested correctness fix with no behavioural regressions.
The change is small and focused: it adds the previously-missing check_gas_limit call to the parallel receipt-building path, matches the logic of the existing sequential path exactly (same argument order, same gas accounting variables), and is covered by a new regression test. No unrelated files are touched and the risk of side effects is minimal.
No files require special attention.

Sequence Diagram

sequenceDiagram
    participant EP as execute_block_parallel
    participant RY as Rayon Tasks
    participant BPR as build_parallel_receipts
    participant CGL as check_gas_limit

    EP->>RY: spawn N tasks with tx_idx 0 to n
    RY->>RY: execute_tx produces ExecutionReport
    RY-->>EP: Vec of (tx_idx, tx_type, tx_gas_limit, report)
    EP->>BPR: build_parallel_receipts(block_gas_limit, results)
    BPR->>BPR: sort results by tx_idx
    loop for each tx in order
        BPR->>CGL: check_gas_limit(cumulative_gas_used, tx_gas_limit, block_gas_limit)
        CGL-->>BPR: Ok or EvmError
        BPR->>BPR: cumulative_gas_used += gas_spent
        BPR->>BPR: block_gas_used += gas_used
        BPR->>BPR: push Receipt
    end
    BPR-->>EP: receipts and block_gas_used

_{Last reviewed commit: 8b69e7f}

edg-l

Good catch! Thanks for the PR

The CI does it but I validated the ef-tests still pass with
make -C tooling/ef_tests/blockchain test

I checked if geth does this check, but it seems they don't either, so they are probs missing this validation too.

ElFantasma · 2026-03-08T22:18:16Z

+    let mut block_gas_used = 0_u64;
+
+    for (_, tx_type, tx_gas_limit, report) in results {
+        check_gas_limit(cumulative_gas_used, tx_gas_limit, block_gas_limit)?;


Bug: this passes cumulative_gas_used (sum of gas_spent, post-refund) to check_gas_limit, but the sequential paths (lines 107 and 215) pass block_gas_used (sum of gas_used, pre-refund). Post-EIP-7778 these differ: gas_spent <= gas_used, so the parallel path becomes more lenient and could accept blocks that the sequential path rejects.

Should be:

check_gas_limit(block_gas_used, tx_gas_limit, block_gas_limit)?;

Sorry, I got this wrong. The sequential paths (lines 122 and 285 on main) actually use cumulative_gas_used (post-refund), not block_gas_used. PR #6309 explicitly fixed this to use post-refund gas because pre-refund gas causes false "Gas allowance exceeded" rejections on tightly-packed Amsterdam blocks with large SSTORE refunds (see the post-mortem there).

Your original code using cumulative_gas_used was correct. Commit 6b7f80 switching to block_gas_used reintroduces the bug that #6309 fixed. Please revert to cumulative_gas_used for the parallel path to match the sequential paths.

ElFantasma · 2026-03-08T22:18:16Z

    }
+
+    #[test]
+    fn test_build_parallel_receipts_rejects_tx_that_exceeds_remaining_block_gas() {


nit: the test uses gas_used == gas_spent for all entries, so it doesn't exercise the bug described above. Consider using different values (e.g., gas_used: 70, gas_spent: 60) to verify the check uses the pre-refund value.

Disregard this too — since the check should use cumulative_gas_used (post-refund), the test values should match that. Though having gas_used != gas_spent in the test is still useful to verify the receipt uses gas_spent for cumulative_gas_used correctly.

Thanks so much! I have been following the project for a while, so I'm really happy I could contribute!!

ElFantasma

Approved

Sorry again for my mistake, and thanks for your contribution!

ilitteri · 2026-03-11T16:06:11Z

Hey @sueun-dev, the PR is good to go. At ethrex, we have a commit signature verification policy that states that all commits must be verified (read here).

Please, sign the commits of this PR for it to be able to be merged.

sueun-dev · 2026-03-11T16:39:03Z

…lass#6342) ## Summary - carry each transaction's `gas_limit()` into the BAL-backed parallel execution result set - enforce the same remaining block gas check used by the sequential LEVM path when building parallel receipts - add a regression test for a later transaction whose `gas_limit` exceeds the remaining block gas Fixes lambdaclass#6341. ## Testing - `cargo test -p ethrex-vm` Co-authored-by: Sueun Cho <sueun-dev@users.noreply.github.com> Co-authored-by: ElFantasma <estebandh@gmail.com>

Copilot AI review requested due to automatic review settings March 8, 2026 20:52

sueun-dev requested a review from a team as a code owner March 8, 2026 20:52

Copilot started reviewing on behalf of sueun-dev March 8, 2026 20:53 View session

Copilot AI reviewed Mar 8, 2026

View reviewed changes

sueun-dev changed the title ~~Validate remaining block gas in parallel LEVM pipeline~~ fix(levm): validate remaining block gas in parallel pipeline Mar 8, 2026

edg-l approved these changes Mar 9, 2026

View reviewed changes

ElFantasma requested changes Mar 9, 2026

View reviewed changes

ElFantasma approved these changes Mar 9, 2026

View reviewed changes

ilitteri approved these changes Mar 11, 2026

View reviewed changes

ilitteri enabled auto-merge March 11, 2026 13:46

fix(levm): validate remaining block gas in parallel pipeline

6fcaf4c

auto-merge was automatically disabled March 11, 2026 16:36
Head branch was pushed to by a user without write access

sueun-dev force-pushed the fix-parallel-gas-limit-check branch from 67660a0 to 6fcaf4c Compare March 11, 2026 16:36

ElFantasma mentioned this pull request Mar 12, 2026

feat(l1): bal-devnet-3 #6216

Merged

11 tasks

Merge branch 'main' into fix-parallel-gas-limit-check

f273927

ElFantasma added this pull request to the merge queue Mar 16, 2026

Merged via the queue into lambdaclass:main with commit 5e9f245 Mar 16, 2026
58 checks passed

Sunbelt Computer Software

PL/B Language Development and Support

Uh oh!

Conversation

sueun-dev commented Mar 8, 2026

Summary

Testing

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

greptile-apps Bot commented Mar 8, 2026

Greptile Summary

Confidence Score: 5/5

Sequence Diagram

Uh oh!

edg-l left a comment

Choose a reason for hiding this comment

Uh oh!

ElFantasma Mar 8, 2026

Choose a reason for hiding this comment

Uh oh!

sueun-dev Mar 9, 2026

Choose a reason for hiding this comment

Uh oh!

ElFantasma Mar 9, 2026

Choose a reason for hiding this comment

Uh oh!

ElFantasma Mar 8, 2026

Choose a reason for hiding this comment

Uh oh!

ElFantasma Mar 9, 2026

Choose a reason for hiding this comment

Uh oh!

sueun-dev Mar 9, 2026

Choose a reason for hiding this comment

Uh oh!

ElFantasma left a comment

Choose a reason for hiding this comment

Uh oh!

ilitteri commented Mar 11, 2026

Uh oh!

sueun-dev commented Mar 11, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants