OP Stack Specification

Table of Contents

• About Optimism
• About the OP Stack
• Site Navigation

About Optimism

Optimism is a project dedicated to scaling Ethereum's technology and expanding its ability to coordinate people from across the world to build effective decentralized economies and governance systems. The Optimism Collective builds open-source software that powers scalable blockchains and aims to address key governance and economic challenges in the wider Ethereum ecosystem. Optimism operates on the principle of impact=profit, the idea that individuals who positively impact the Collective should be proportionally rewarded with profit.

Change the incentives and you change the world.

About the OP Stack

The OP Stack is a decentralized software stack maintained by the OP Stack that forms the backbone of blockchains like OP Mainnet and Base. The OP Stack is designed to be aggressively open-source — you are welcome to explore, modify, and extend the OP Stack to your heart's content.

Site Navigation

Navigate this site using the sidebar on the left, the search icon found at the top of this page, or the left/right navigation buttons found to the sides of each page.

Background

Table of Contents

• Overview
• Foundations
  • Ethereum Scalability
  • Optimistic Rollups
  • EVM Equivalence
• Protocol Guarantees
  • Liveness
  • Validity
  • Availability
• Network Participants
  • Users
  • Sequencers
  • Verifiers
• Key Interaction Diagrams
  • Depositing and Sending Transactions
  • Withdrawing
• Next Steps

Overview

The OP Stack is a decentralized software stack maintained by the OP Stack that forms the backbone of blockchains like OP Mainnet and Base. The OP Stack provides the infrastructure for operating EVM equivalent rollup blockchains designed to scale Ethereum while remaining maximally compatible with existing Ethereum infrastructure. This document provides an overview of the protocol to provide context for the rest of the specification.

Foundations

Ethereum Scalability

Scaling Ethereum means increasing the number of useful transactions the Ethereum network can process. Ethereum's limited resources, specifically bandwidth, computation, and storage, constrain the number of transactions which can be processed on the network. Of the three resources, computation and storage are currently the most significant bottlenecks. These bottlenecks limit the supply of transactions, leading to extremely high fees. Scaling ethereum and reducing fees can be achieved by better utilizing bandwidth, computation and storage.

Optimistic Rollups

An Optimistic Rollup is a layer 2 scalability construction which increases the computation & storage capacity of Ethereum while aiming to minimize sacrifices to scalability or decentralization. In a nutshell, an Optimistic Rollup utilizes Ethereum (or some other data availability layer) to host transaction data. Layer 2 nodes then execute a state transition function over this data. Users can propose the result of this off-chain execution to a smart contract on L1. A "fault proving" process can then demonstrate that a user's proposal is (or is not) valid.

EVM Equivalence

EVM Equivalence is complete compliance with the state transition function described in the Ethereum yellow paper, the formal definition of the protocol. By conforming to the Ethereum standard across EVM equivalent rollups, smart contract developers can write once and deploy anywhere.

Protocol Guarantees

We strive to preserve three critical properties: liveness, validity, and availability. A protocol that can maintain these properties can, effectively, scale Ethereum without sacrificing security.

Liveness

Liveness is defined as the ability for any party to be able to extend the rollup chain by including a transaction within a bounded amount of time. It should not be possible for an actor to block the inclusion of any given transaction for more than this bounded time period. This bounded time period should also be acceptable such that inclusion is not just theoretically possible but practically useful.

Validity

Validity is defined as the ability for any party to execute the rollup state transition function, subject to certain lower bound expectations for available computing and bandwidth resources. Validity is also extended to refer to the ability for a smart contract on Ethereum to be able to validate this state transition function economically.

Availability

Availability is defined as the ability for any party to retrieve the inputs that are necessary to execute the rollup state transition function correctly. Availability is essentially an element of validity and is required to be able to guarantee validity in general. Similar to validity, availability is subject to lower bound resource requirements.

Network Participants

Generally speaking, there are three primary actors that interact with an OP Stack chain: users, sequencers, and verifiers.

graph TD
    EthereumL1(Ethereum L1)

    subgraph "L2 Participants"
        Users(Users)
        Sequencers(Sequencers)
        Verifiers(Verifiers)
    end

    Verifiers -.->|fetch transaction batches| EthereumL1
    Verifiers -.->|fetch deposit data| EthereumL1
    Verifiers -->|submit/validate/challenge output proposals| EthereumL1
    Verifiers -.->|fetch realtime P2P updates| Sequencers
    
    Users -->|submit deposits/withdrawals| EthereumL1
    Users -->|submit transactions| Sequencers
    Users -->|query data| Verifiers


    Sequencers -->|submit transaction batches| EthereumL1
    Sequencers -.->|fetch deposit data| EthereumL1

    classDef l1Contracts stroke:#bbf,stroke-width:2px;
    classDef l2Components stroke:#333,stroke-width:2px;
    classDef systemUser stroke:#f9a,stroke-width:2px;

    class EthereumL1 l1Contracts;
    class Users,Sequencers,Verifiers l2Components;

Users

Users are the general class of network participants who:

• Submit transactions through a Sequencer or by interacting with contracts on Ethereum.
• Query transaction data from interfaces operated by verifiers.

Sequencers

Sequencers fill the role of the block producer on an OP Stack chain. Chains may have a single Sequencer or may choose to utilize some consensus protocol that coordinates multiple Sequencers. The OP Stack currently officially only supports a single active Sequencer at any given time. In general, specifications may use the term "the Sequencer" as a stand-in for either a single Sequencer or a consensus protocol of multiple Sequencers.

The Sequencer:

• Accepts transactions directly from Users.
• Observes "deposit" transactions generated on Ethereum.
• Consolidates both transaction streams into ordered L2 blocks.
• Submits information to L1 that is sufficient to fully reproduce those L2 blocks.
• Provides real-time access to pending L2 blocks that have not yet been confirmed on L1.

The Sequencer serves an important role for the operation of an L2 chain but is not a trusted actor. The Sequencer is generally responsible for improving the user experience by ordering transactions much more quickly and cheaply than would currently be possible if users were to submit all transactions directly to L1.

Verifiers

Verifiers download and execute the L2 state transition function independently of the Sequencer. Verifiers help to maintain the integrity of the network and serve blockchain data to Users.

Verifiers generally:

• Download rollup data from L1 and the Sequencer.
• Use rollup data to execute the L2 state transition function.
• Serve rollup data and computed L2 state information to Users.

Verifiers can also act as Proposers and/or Challengers who:

• Submit assertions about the state of the L2 to a smart contract on L1.
• Validate assertions made by other participants.
• Dispute invalid assertions made by other participants.

Key Interaction Diagrams

Depositing and Sending Transactions

Users will often begin their L2 journey by depositing ETH from L1. Once they have ETH to pay fees, they'll start sending transactions on L2. The following diagram demonstrates this interaction and all key OP Stack components which are or should be utilized:

graph TD
    subgraph "Ethereum L1"
        OptimismPortal(<a href="./protocol/withdrawals.html#the-optimism-portal-contract">OptimismPortal</a>)
        BatchInbox(<a href="../glossary.html#batcher-transaction">Batch Inbox Address</a>)
    end

    Sequencer(Sequencer)
    Users(Users)

    %% Interactions
    Users -->|<b>1.</b> submit deposit| OptimismPortal
    Sequencer -.->|<b>2.</b> fetch deposit events| OptimismPortal
    Sequencer -->|<b>3.</b> generate deposit block| Sequencer
    Users -->|<b>4.</b> send transactions| Sequencer
    Sequencer -->|<b>5.</b> submit transaction batches| BatchInbox

    classDef l1Contracts stroke:#bbf,stroke-width:2px;
    classDef l2Components stroke:#333,stroke-width:2px;
    classDef systemUser stroke:#f9a,stroke-width:2px;

    class OptimismPortal,BatchInbox l1Contracts;
    class Sequencer l2Components;
    class Users systemUser;

Withdrawing

Users may also want to withdraw ETH or ERC20 tokens from an OP Stack chain back to Ethereum. Withdrawals are initiated as standard transactions on L2 but are then completed using transactions on L2. Withdrawals must reference a valid FaultDisputeGame contract that proposes the state of the L2 at a given point in time.

graph LR
    subgraph "Ethereum L1"
        BatchInbox(<a href="../glossary.html#batcher-transaction">Batch Inbox Address</a>)
        DisputeGameFactory(<a href="../fault-proof/stage-one/dispute-game-interface.html#disputegamefactory-interface">DisputeGameFactory</a>)
        FaultDisputeGame(<a href="../fault-proof/stage-one/fault-dispute-game.html">FaultDisputeGame</a>)
        OptimismPortal(<a href="./protocol/withdrawals.html#the-optimism-portal-contract">OptimismPortal</a>)
        ExternalContracts(External Contracts)
    end

    Sequencer(Sequencer)
    Proposers(Proposers)
    Users(Users)

    %% Interactions
    Users -->|<b>1.</b> send withdrawal initialization txn| Sequencer
    Sequencer -->|<b>2.</b> submit transaction batch| BatchInbox
    Proposers -->|<b>3.</b> submit output proposal| DisputeGameFactory
    DisputeGameFactory -->|<b>4.</b> generate game| FaultDisputeGame
    Users -->|<b>5.</b> submit withdrawal proof| OptimismPortal
    Users -->|<b>6.</b> wait for finalization| FaultDisputeGame
    Users -->|<b>7.</b> submit withdrawal finalization| OptimismPortal
    OptimismPortal -->|<b>8.</b> check game validity| FaultDisputeGame
    OptimismPortal -->|<b>9.</b> execute withdrawal transaction| ExternalContracts

    %% Styling
    classDef l1Contracts stroke:#bbf,stroke-width:2px;
    classDef l2Components stroke:#333,stroke-width:2px;
    classDef systemUser stroke:#f9a,stroke-width:2px;

    class BatchInbox,DisputeGameFactory,FaultDisputeGame,OptimismPortal l1Contracts;
    class Sequencer l2Components;
    class Users,Proposers systemUser;

Next Steps

Check out the sidebar to the left to find any specification you might want to read, or click one of the links embedded in one of the above diagrams to learn about particular components that have been mentioned.

Optimism Overview

Table of Contents

• Architecture Design Goals
• Architecture Overview
  • Core L1 Smart Contracts
    • Notes for Core L1 Smart Contracts
  • Core L2 Smart Contracts
    • Notes for Core L2 Smart Contracts
  • Smart Contract Proxies
  • L2 Node Components
  • Transaction/Block Propagation
• Key Interactions In Depth
  • Deposits
  • Block Derivation
    • Overview
    • Epochs and the Sequencing Window
    • Block Derivation Loop
  • Engine API

This document is a high-level technical overview of the Optimism protocol. It aims to explain how the protocol works in an informal manner, and direct readers to other parts of the specification so that they may learn more.

This document assumes you've read the background.

Architecture Design Goals

• Execution-Level EVM Equivalence: The developer experience should be identical to L1 except where L2 introduces a fundamental difference.
  • No special compiler.
  • No unexpected gas costs.
  • Transaction traces work out-of-the-box.
  • All existing Ethereum tooling works - all you have to do is change the chain ID.
• Maximal compatibility with ETH1 nodes: The implementation should minimize any differences with a vanilla Geth node, and leverage as many existing L1 standards as possible.
  • The execution engine/rollup node uses the ETH2 Engine API to build the canonical L2 chain.
  • The execution engine leverages Geth's existing mempool and sync implementations, including snap sync.
• Minimize state and complexity:
  • Whenever possible, services contributing to the rollup infrastructure are stateless.
  • Stateful services can recover to full operation from a fresh DB using the peer-to-peer network and on-chain sync mechanisms.
  • Running a replica is as simple as running a Geth node.

Architecture Overview

Core L1 Smart Contracts

Below you'll find an architecture diagram describing the core L1 smart contracts for the OP Stack. Smart contracts that are considered "peripheral" and not core to the operation of the OP Stack system are described separately.

graph LR
    subgraph "External Contracts"
        ExternalERC20(External ERC20 Contracts)
        ExternalERC721(External ERC721 Contracts)
    end

    subgraph "L1 Smart Contracts"
        BatchDataEOA(<a href="../glossary.html#batcher-transaction">Batch Inbox Address</a>)
        L1StandardBridge(<a href="./bridges.html">L1StandardBridge</a>)
        L1ERC721Bridge(<a href="./bridges.html">L1ERC721Bridge</a>)
        L1CrossDomainMessenger(<a href="./messengers.html">L1CrossDomainMessenger</a>)
        OptimismPortal(<a href="./withdrawals.html#the-optimism-portal-contract">OptimismPortal</a>)
        SuperchainConfig(<a href="./superchain-configuration.html">SuperchainConfig</a>)
        SystemConfig(<a href="./system-config.html">SystemConfig</a>)
        DisputeGameFactory(<a href="../fault-proof/stage-one/dispute-game-interface.html#disputegamefactory-interface">DisputeGameFactory</a>)
        FaultDisputeGame(<a href="../fault-proof/stage-one/fault-dispute-game.html">FaultDisputeGame</a>)
        AnchorStateRegistry(<a href="../fault-proof/stage-one/fault-dispute-game.html#anchor-state-registry">AnchorStateRegistry</a>)
        DelayedWETH(<a href="../fault-proof/stage-one/bond-incentives.html#delayedweth#de">DelayedWETH</a>)
    end

    subgraph "User Interactions (Permissionless)"
        Users(Users)
        Challengers(Challengers)
    end

    subgraph "System Interactions"
        Guardian(Guardian)
        Batcher(<a href="./batcher.html">Batcher</a>)
    end

    subgraph "Layer 2 Interactions"
        L2Nodes(Layer 2 Nodes)
    end

    L2Nodes -.->|fetch transaction batches| BatchDataEOA
    L2Nodes -.->|fetch deposit events| OptimismPortal

    Batcher -->|publish transaction batches| BatchDataEOA

    ExternalERC20 <-->|mint/burn/transfer tokens| L1StandardBridge
    ExternalERC721 <-->|mint/burn/transfer tokens| L1ERC721Bridge

    L1StandardBridge <-->|send/receive messages| L1CrossDomainMessenger
    L1StandardBridge -.->|query pause state| SuperchainConfig

    L1ERC721Bridge <-->|send/receive messages| L1CrossDomainMessenger
    L1ERC721Bridge -.->|query pause state| SuperchainConfig

    L1CrossDomainMessenger <-->|send/receive messages| OptimismPortal
    L1CrossDomainMessenger -.->|query pause state| SuperchainConfig

    OptimismPortal -.->|query pause state| SuperchainConfig
    OptimismPortal -.->|query config| SystemConfig
    OptimismPortal -.->|query state proposals| DisputeGameFactory

    DisputeGameFactory -->|generate instances| FaultDisputeGame

    FaultDisputeGame -->|store bonds| DelayedWETH
    FaultDisputeGame -->|query/update anchor states| AnchorStateRegistry

    Users <-->|deposit/withdraw ETH/ERC20s| L1StandardBridge
    Users <-->|deposit/withdraw ERC721s| L1ERC721Bridge
    Users -->|prove/execute withdrawals| OptimismPortal

    Challengers -->|propose output roots| DisputeGameFactory
    Challengers -->|verify/challenge/defend proposals| FaultDisputeGame

    Guardian -->|pause/unpause| SuperchainConfig
    Guardian -->|safety net actions| OptimismPortal
    Guardian -->|safety net actions| DisputeGameFactory
    Guardian -->|safety net actions| DelayedWETH

    classDef extContracts stroke:#ff9,stroke-width:2px;
    classDef l1Contracts stroke:#bbf,stroke-width:2px;
    classDef l1EOA stroke:#bbb,stroke-width:2px;
    classDef userInt stroke:#f9a,stroke-width:2px;
    classDef systemUser stroke:#f9a,stroke-width:2px;
    classDef l2Nodes stroke:#333,stroke-width:2px
    class ExternalERC20,ExternalERC721 extContracts;
    class L1StandardBridge,L1ERC721Bridge,L1CrossDomainMessenger,OptimismPortal,SuperchainConfig,SystemConfig,DisputeGameFactory,FaultDisputeGame,DelayedWETH,AnchorStateRegistry l1Contracts;
    class BatchDataEOA l1EOA;
    class Users,Challengers userInt;
    class Batcher,Guardian systemUser;
    class L2Nodes l2Nodes;

Notes for Core L1 Smart Contracts

• The Batch Inbox Address described above (highlighted in GREY) is not a smart contract and is instead an arbitrarily selected account that is assumed to have no known private key. The convention for deriving this account's address is provided on the Configurability page.
  • Historically, it was often derived as 0xFF0000....<L2 chain ID> where <L2 chain ID> is chain ID of the Layer 2 network for which the data is being posted. This is why many chains, such as OP Mainnet, have a batch inbox address of this form.
• Smart contracts that sit behind Proxy contracts are highlighted in BLUE. Refer to the Smart Contract Proxies section below to understand how these proxies are designed.
  • The L1CrossDomainMessenger contract sits behind the ResolvedDelegateProxy contract, a legacy proxy contract type used within older versions of the OP Stack. This proxy type is used exclusively for the L1CrossDomainMessenger to maintain backwards compatibility.
  • The L1StandardBridge contract sits behind the L1ChugSplashProxy contract, a legacy proxy contract type used within older versions of the OP Stack. This proxy type is used exclusively for the L1StandardBridge contract to maintain backwards compatibility.

Core L2 Smart Contracts

Here you'll find an architecture diagram describing the core OP Stack smart contracts that exist natively on the L2 chain itself.

graph LR
    subgraph "Layer 1 (Ethereum)"
        L1SmartContracts(L1 Smart Contracts)
    end

    subgraph "L2 Client"
        L2Node(L2 Node)
    end

    subgraph "L2 System Contracts"
        L1Block(<a href="./predeploys.html#l1block">L1Block</a>)
        GasPriceOracle(<a href="./predeploys.html#gaspriceoracle">GasPriceOracle</a>)
        L1FeeVault(<a href="./predeploys.html#l1feevault">L1FeeVault</a>)
        BaseFeeVault(<a href="./predeploys.html#basefeevault">BaseFeeVault</a>)
        SequencerFeeVault(<a href="./predeploys.html#sequencerfeevault">SequencerFeeVault</a>)
    end

    subgraph "L2 Bridge Contracts"
        L2CrossDomainMessenger(<a href="./predeploys.html#l2crossdomainmessenger">L2CrossDomainMessenger</a>)
        L2ToL1MessagePasser(<a href="./predeploys.html#l2tol1messagepasser">L2ToL1MessagePasser</a>)
        L2StandardBridge(<a href="./predeploys.html#l2standardbridge">L2StandardBridge</a>)
        L2ERC721Bridge(<a href="./predeploys.html">L2ERC721Bridge</a>)
    end

    subgraph "Transactions"
        DepositTransaction(Deposit Transaction)
        UserTransaction(User Transaction)
    end

    subgraph "External Contracts"
        ExternalERC20(External ERC20 Contracts)
        ExternalERC721(External ERC721 Contracts)
    end

    subgraph "Remaining L2 Universe"
        OtherContracts(Any Contracts and Addresses)
    end

    L2Node -.->|derives chain from| L1SmartContracts
    L2Node -->|updates| L1Block
    L2Node -->|distributes fees to| L1FeeVault
    L2Node -->|distributes fees to| BaseFeeVault
    L2Node -->|distributes fees to| SequencerFeeVault
    L2Node -->|derives from deposits| DepositTransaction
    L2Node -->|derives from chain data| UserTransaction

    UserTransaction -->|can trigger| OtherContracts
    DepositTransaction -->|maybe triggers| L2CrossDomainMessenger
    DepositTransaction -->|can trigger| OtherContracts

    ExternalERC20 <-->|mint/burn/transfer| L2StandardBridge
    ExternalERC721 <-->|mint/burn/transfer| L2ERC721Bridge

    L2StandardBridge <-->|sends/receives messages| L2CrossDomainMessenger
    L2ERC721Bridge <-->|sends/receives messages| L2CrossDomainMessenger
    GasPriceOracle -.->|queries| L1Block
    L2CrossDomainMessenger -->|sends messages| L2ToL1MessagePasser

    classDef extContracts stroke:#ff9,stroke-width:2px;
    classDef l2Contracts stroke:#bbf,stroke-width:2px;
    classDef transactions stroke:#fba,stroke-width:2px;
    classDef l2Node stroke:#f9a,stroke-width:2px;

    class ExternalERC20,ExternalERC721 extContracts;
    class L2CrossDomainMessenger,L2ToL1MessagePasser,L2StandardBridge,L2ERC721Bridge l2Contracts;
    class L1Block,L1FeeVault,BaseFeeVault,SequencerFeeVault,GasPriceOracle l2Contracts;
    class UserTransaction,DepositTransaction transactions;
    class L2Node l2Node;

Notes for Core L2 Smart Contracts

• Contracts highlighted as "L2 System Contracts" are updated or mutated automatically as part of the chain derivation process. Users typically do not mutate these contracts directly, except in the case of the FeeVault contracts where any user may trigger a withdrawal of collected fees to the pre-determined withdrawal address.
• Smart contracts that sit behind Proxy contracts are highlighted in BLUE. Refer to the Smart Contract Proxies section below to understand how these proxies are designed.
• User interactions for the "L2 Bridge Contracts" have been omitted from this diagram but largely follow the same user interactions described in the architecture diagram for the Core L1 Smart Contracts.

Smart Contract Proxies

Most OP Stack smart contracts sit behind Proxy contracts that are managed by a ProxyAdmin contract. The ProxyAdmin contract is controlled by some owner address that can be any EOA or smart contract. Below you'll find a diagram that explains the behavior of the typical proxy contract.

graph LR
    ProxyAdminOwner(Proxy Admin Owner)
    ProxyAdmin(<a href="https://github.com/ethereum-optimism/optimism/blob/develop/packages/contracts-bedrock/src/universal/ProxyAdmin.sol">ProxyAdmin</a>)

    subgraph "Logical Smart Contract"
        Proxy(<a href="https://github.com/ethereum-optimism/optimism/blob/develop/packages/contracts-bedrock/src/universal/Proxy.sol">Proxy</a>)
        Implementation(Implementation)
    end

    ProxyAdminOwner -->|manages| ProxyAdmin
    ProxyAdmin -->|upgrades| Proxy
    Proxy -->|delegatecall| Implementation

    classDef l1Contracts stroke:#bbf,stroke-width:2px;
    classDef systemUser stroke:#f9a,stroke-width:2px;
    class Proxy l1Contracts;
    class ProxyAdminOwner systemUser;

L2 Node Components

Below you'll find a diagram illustrating the basic interactions between the components that make up an L2 node as well as demonstrations of how different actors use these components to fulfill their roles.

graph LR
    subgraph "L2 Node"
        RollupNode(<a href="./rollup-node.html">Rollup Node</a>)
        ExecutionEngine(<a href="./exec-engine.html">Execution Engine</a>)
    end

    subgraph "System Interactions"
        BatchSubmitter(<a href="./batcher.html">Batch Submitter</a>)
        OutputSubmitter(Output Submitter)
        Challenger(Challenger)
    end

    subgraph "L1 Smart Contracts"
        BatchDataEOA(<a href="../glossary.html#batcher-transaction">Batch Inbox Address</a>)
        OptimismPortal(<a href="./withdrawals.html#the-optimism-portal-contract">OptimismPortal</a>)
        DisputeGameFactory(<a href="../fault-proof/stage-one/dispute-game-interface.html#disputegamefactory-interface">DisputeGameFactory</a>)
        FaultDisputeGame(<a href="../fault-proof/stage-one/fault-dispute-game.html">FaultDisputeGame</a>)
    end

    BatchSubmitter -.->|fetch transaction batch info| RollupNode
    BatchSubmitter -.->|fetch transaction batch info| ExecutionEngine
    BatchSubmitter -->|send transaction batches| BatchDataEOA

    RollupNode -.->|fetch transaction batches| BatchDataEOA
    RollupNode -.->|fetch deposit transactions| OptimismPortal
    RollupNode -->|drives| ExecutionEngine

    OutputSubmitter -.->|fetch outputs| RollupNode
    OutputSubmitter -->|send output proposals| DisputeGameFactory

    Challenger -.->|fetch dispute games| DisputeGameFactory
    Challenger -->|verify/challenge/defend games| FaultDisputeGame

    classDef l2Components stroke:#333,stroke-width:2px;
    classDef systemUser stroke:#f9a,stroke-width:2px;
    classDef l1Contracts stroke:#bbf,stroke-width:2px;

    class RollupNode,ExecutionEngine l2Components;
    class BatchSubmitter,OutputSubmitter,Challenger systemUser;
    class BatchDataEOA,OptimismPortal,DisputeGameFactory,FaultDisputeGame l1Contracts;

Transaction/Block Propagation

Spec links:

• Execution Engine

Since the EE uses Geth under the hood, Optimism uses Geth's built-in peer-to-peer network and transaction pool to propagate transactions. The same network can also be used to propagate submitted blocks and support snap-sync.

Unsubmitted blocks, however, are propagated using a separate peer-to-peer network of Rollup Nodes. This is optional, however, and is provided as a convenience to lower latency for verifiers and their JSON-RPC clients.

The below diagram illustrates how the sequencer and verifiers fit together:

Key Interactions In Depth

Deposits

Spec links:

• Deposits

Optimism supports two types of deposits: user deposits, and L1 attributes deposits. To perform a user deposit, users call the depositTransaction method on the OptimismPortal contract. This in turn emits TransactionDeposited events, which the rollup node reads during block derivation.

L1 attributes deposits are used to register L1 block attributes (number, timestamp, etc.) on L2 via a call to the L1 Attributes Predeploy. They cannot be initiated by users, and are instead added to L2 blocks automatically by the rollup node.

Both deposit types are represented by a single custom EIP-2718 transaction type on L2.

Block Derivation

Overview

The rollup chain can be deterministically derived given an L1 Ethereum chain. The fact that the entire rollup chain can be derived based on L1 blocks is what makes Optimism a rollup. This process can be represented as:

derive_rollup_chain(l1_blockchain) -> rollup_blockchain

Optimism's block derivation function is designed such that it:

• Requires no state other than what is easily accessible using L1 and L2 execution engine APIs.
• Supports sequencers and sequencer consensus.
• Is resilient to sequencer censorship.

Epochs and the Sequencing Window

The rollup chain is subdivided into epochs. There is a 1:1 correspondence between L1 block numbers and epoch numbers.

For L1 block number n, there is a corresponding rollup epoch n which can only be derived after a sequencing window worth of blocks has passed, i.e. after L1 block number n + SEQUENCING_WINDOW_SIZE is added to the L1 chain.

Each epoch contains at least one block. Every block in the epoch contains an L1 info transaction which contains contextual information about L1 such as the block hash and timestamp. The first block in the epoch also contains all deposits initiated via the OptimismPortal contract on L1. All L2 blocks can also contain sequenced transactions, i.e. transactions submitted directly to the sequencer.

Whenever the sequencer creates a new L2 block for a given epoch, it must submit it to L1 as part of a batch, within the epoch's sequencing window (i.e. the batch must land before L1 block n + SEQUENCING_WINDOW_SIZE). These batches are (along with the TransactionDeposited L1 events) what allows the derivation of the L2 chain from the L1 chain.

The sequencer does not need for a L2 block to be batch-submitted to L1 in order to build on top of it. In fact, batches typically contain multiple L2 blocks worth of sequenced transactions. This is what enables fast transaction confirmations on the sequencer.

Since transaction batches for a given epoch can be submitted anywhere within the sequencing window, verifiers must search all blocks within the window for transaction batches. This protects against the uncertainty of transaction inclusion of L1. This uncertainty is also why we need the sequencing window in the first place: otherwise the sequencer could retroactively add blocks to an old epoch, and validators wouldn't know when they can finalize an epoch.

The sequencing window also prevents censorship by the sequencer: deposits made on a given L1 block will be included in the L2 chain at worst after SEQUENCING_WINDOW_SIZE L1 blocks have passed.

The following diagram describes this relationship, and how L2 blocks are derived from L1 blocks (L1 info transactions have been elided):

Block Derivation Loop

A sub-component of the rollup node called the rollup driver is actually responsible for performing block derivation. The rollup driver is essentially an infinite loop that runs the block derivation function. For each epoch, the block derivation function performs the following steps:

1. Downloads deposit and transaction batch data for each block in the sequencing window.
2. Converts the deposit and transaction batch data into payload attributes for the Engine API.
3. Submits the payload attributes to the Engine API, where they are converted into blocks and added to the canonical chain.

This process is then repeated with incrementing epochs until the tip of L1 is reached.

Engine API

The rollup driver doesn't actually create blocks. Instead, it directs the execution engine to do so via the Engine API. For each iteration of the block derivation loop described above, the rollup driver will craft a payload attributes object and send it to the execution engine. The execution engine will then convert the payload attributes object into a block, and add it to the chain. The basic sequence of the rollup driver is as follows:

1. Call fork choice updated with the payload attributes object. We'll skip over the details of the fork choice state parameter for now - just know that one of its fields is the L2 chain's headBlockHash, and that it is set to the block hash of the tip of the L2 chain. The Engine API returns a payload ID.
2. Call get payload with the payload ID returned in step 1. The engine API returns a payload object that includes a block hash as one of its fields.
3. Call new payload with the payload returned in step 2. (Ecotone blocks, must use V3, pre-Ecotone blocks MUST use the V2 version)
4. Call fork choice updated with the fork choice parameter's headBlockHash set to the block hash returned in step 2. The tip of the L2 chain is now the block created in step 1.

The swimlane diagram below visualizes the process:

Standard Bridges

Table of Contents

• Overview
• Token Depositing
• Upgradability

Overview

The standard bridges are responsible for allowing cross domain ETH and ERC20 token transfers. They are built on top of the cross domain messenger contracts and give a standard interface for depositing tokens.

The bridge works for both L1 native tokens and L2 native tokens. The legacy API is preserved to ensure that existing applications will not experience any problems with the Bedrock StandardBridge contracts.

The L2StandardBridge is a predeploy contract located at 0x4200000000000000000000000000000000000010.

interface StandardBridge {
    event ERC20BridgeFinalized(address indexed localToken, address indexed remoteToken, address indexed from, address to, uint256 amount, bytes extraData);
    event ERC20BridgeInitiated(address indexed localToken, address indexed remoteToken, address indexed from, address to, uint256 amount, bytes extraData);
    event ETHBridgeFinalized(address indexed from, address indexed to, uint256 amount, bytes extraData);
    event ETHBridgeInitiated(address indexed from, address indexed to, uint256 amount, bytes extraData);

    function bridgeERC20(address _localToken, address _remoteToken, uint256 _amount, uint32 _minGasLimit, bytes memory _extraData) external;
    function bridgeERC20To(address _localToken, address _remoteToken, address _to, uint256 _amount, uint32 _minGasLimit, bytes memory _extraData) external;
    function bridgeETH(uint32 _minGasLimit, bytes memory _extraData) payable external;
    function bridgeETHTo(address _to, uint32 _minGasLimit, bytes memory _extraData) payable external;
    function deposits(address, address) view external returns (uint256);
    function finalizeBridgeERC20(address _localToken, address _remoteToken, address _from, address _to, uint256 _amount, bytes memory _extraData) external;
    function finalizeBridgeETH(address _from, address _to, uint256 _amount, bytes memory _extraData) payable external;
    function messenger() view external returns (address);
    function OTHER_BRIDGE() view external returns (address);
}

Token Depositing

The bridgeERC20 function is used to send a token from one domain to another domain. An OptimismMintableERC20 token contract must exist on the remote domain to be able to deposit tokens to that domain. One of these tokens can be deployed using the OptimismMintableERC20Factory contract.

Upgradability

Both the L1 and L2 standard bridges should be behind upgradable proxies.

Cross Domain Messengers

Table of Contents

• Overview
• Message Passing
• Upgradability
• Message Versioning
  • Message Version 0
  • Message Version 1
• Backwards Compatibility Notes

Overview

The cross domain messengers are responsible for providing a higher level API for developers who are interested in sending cross domain messages. They allow for the ability to replay cross domain messages and sit directly on top of the lower level system contracts responsible for cross domain messaging on L1 and L2.

The CrossDomainMessenger is extended to create both an L1CrossDomainMessenger as well as a L2CrossDomainMessenger. These contracts are then extended with their legacy APIs to provide backwards compatibility for applications that integrated before the Bedrock system upgrade.

The L2CrossDomainMessenger is a predeploy contract located at 0x4200000000000000000000000000000000000007.

The base CrossDomainMessenger interface is:

interface CrossDomainMessenger {
    event FailedRelayedMessage(bytes32 indexed msgHash);
    event RelayedMessage(bytes32 indexed msgHash);
    event SentMessage(address indexed target, address sender, bytes message, uint256 messageNonce, uint256 gasLimit);
    event SentMessageExtension1(address indexed sender, uint256 value);

    function MESSAGE_VERSION() external view returns (uint16);
    function MIN_GAS_CALLDATA_OVERHEAD() external view returns (uint64);
    function MIN_GAS_CONSTANT_OVERHEAD() external view returns (uint64);
    function MIN_GAS_DYNAMIC_OVERHEAD_DENOMINATOR() external view returns (uint64);
    function MIN_GAS_DYNAMIC_OVERHEAD_NUMERATOR() external view returns (uint64);
    function OTHER_MESSENGER() external view returns (address);
    function baseGas(bytes memory _message, uint32 _minGasLimit) external pure returns (uint64);
    function failedMessages(bytes32) external view returns (bool);
    function messageNonce() external view returns (uint256);
    function relayMessage(
        uint256 _nonce,
        address _sender,
        address _target,
        uint256 _value,
        uint256 _minGasLimit,
        bytes memory _message
    ) external payable returns (bytes memory returnData_);
    function sendMessage(address _target, bytes memory _message, uint32 _minGasLimit) external payable;
    function successfulMessages(bytes32) external view returns (bool);
    function xDomainMessageSender() external view returns (address);
}

Message Passing

The sendMessage function is used to send a cross domain message. To trigger the execution on the other side, the relayMessage function is called. Successful messages have their hash stored in the successfulMessages mapping while unsuccessful messages have their hash stored in the failedMessages mapping.

The user experience when sending from L1 to L2 is a bit different than when sending a transaction from L2 to L1. When going into L1 from L2, the user does not need to call relayMessage on L2 themselves. The user pays for L2 gas on L1 and the transaction is automatically pulled into L2 where it is executed on L2. When going from L2 into L1, the user proves their withdrawal on OptimismPortal, then waits for the finalization window to pass, and then finalizes the withdrawal on the OptimismPortal, which calls relayMessage on the L1CrossDomainMessenger to finalize the withdrawal.

Upgradability

The L1 and L2 cross domain messengers should be deployed behind upgradable proxies. This will allow for updating the message version.

Message Versioning

Messages are versioned based on the first 2 bytes of their nonce. Depending on the version, messages can have a different serialization and hashing scheme. The first two bytes of the nonce are reserved for version metadata because a version field was not originally included in the messages themselves, but a uint256 nonce is so large that we can very easily pack additional data into that field.

Message Version 0

abi.encodeWithSignature(
    "relayMessage(address,address,bytes,uint256)",
    _target,
    _sender,
    _message,
    _messageNonce
);

Message Version 1

abi.encodeWithSignature(
    "relayMessage(uint256,address,address,uint256,uint256,bytes)",
    _nonce,
    _sender,
    _target,
    _value,
    _gasLimit,
    _data
);

Backwards Compatibility Notes

An older version of the messenger contracts had the concept of blocked messages in a blockedMessages mapping. This functionality was removed from the messengers because a smart attacker could get around any message blocking attempts. It also saves gas on finalizing withdrawals.

The concept of a "relay id" and the relayedMessages mapping was removed. It was built as a way to be able to fund third parties who relayed messages on the behalf of users, but it was improperly implemented as it was impossible to know if the relayed message actually succeeded.

Deposits

Table of Contents

• Overview
• The Deposited Transaction Type
  • Source hash computation
  • Kinds of Deposited Transactions
  • Validation and Authorization of Deposited Transactions
  • Execution
    • Nonce Handling
• Deposit Receipt
• L1 Attributes Deposited Transaction
  • L1 Attributes Deposited Transaction Calldata
    • L1 Attributes - Bedrock, Canyon, Delta
• Special Accounts on L2
  • L1 Attributes Depositor Account
  • L1 Attributes Predeployed Contract
    • L1 Attributes Predeployed Contract: Reference Implementation
• User-Deposited Transactions
  • Deposit Contract
    • Address Aliasing
    • Deposit Contract Implementation: Optimism Portal

Overview

Deposited transactions, also known as deposits are transactions which are initiated on L1, and executed on L2. This document outlines a new transaction type for deposits. It also describes how deposits are initiated on L1, along with the authorization and validation conditions on L2.

Vocabulary note: deposited transaction refers specifically to an L2 transaction, while deposit can refer to the transaction at various stages (for instance when it is deposited on L1).

The Deposited Transaction Type

Deposited transactions have the following notable distinctions from existing transaction types:

1. They are derived from Layer 1 blocks, and must be included as part of the protocol.
2. They do not include signature validation (see User-Deposited Transactions for the rationale).
3. They buy their L2 gas on L1 and, as such, the L2 gas is not refundable.

We define a new EIP-2718 compatible transaction type with the prefix 0x7E to represent a deposit transaction.

A deposit has the following fields (rlp encoded in the order they appear here):

• bytes32 sourceHash: the source-hash, uniquely identifies the origin of the deposit.
• address from: The address of the sender account.
• address to: The address of the recipient account, or the null (zero-length) address if the deposited transaction is a contract creation.
• uint256 mint: The ETH value to mint on L2.
• uint256 value: The ETH value to send to the recipient account.
• uint64 gas: The gas limit for the L2 transaction.
• bool isSystemTx: If true, the transaction does not interact with the L2 block gas pool.
  • Note: boolean is disabled (enforced to be false) starting from the Regolith upgrade.
• bytes data: The calldata.

In contrast to EIP-155 transactions, this transaction type:

• Does not include a nonce, since it is identified by the sourceHash. API responses still include a nonce attribute:
  • Before Regolith: the nonce is always 0
  • With Regolith: the nonce is set to the depositNonce attribute of the corresponding transaction receipt.
• Does not include signature information, and makes the from address explicit. API responses contain zeroed signature v, r, s values for backwards compatibility.
• Includes new sourceHash, from, mint, and isSystemTx attributes. API responses contain these as additional fields.

We select 0x7E because transaction type identifiers are currently allowed to go up to 0x7F. Picking a high identifier minimizes the risk that the identifier will be used by another transaction type on the L1 chain in the future. We don't pick 0x7F itself in case it becomes used for a variable-length encoding scheme.

Source hash computation

The sourceHash of a deposit transaction is computed based on the origin:

• User-deposited: keccak256(bytes32(uint256(0)), keccak256(l1BlockHash, bytes32(uint256(l1LogIndex)))). Where the l1BlockHash, and l1LogIndex all refer to the inclusion of the deposit log event on L1. l1LogIndex is the index of the deposit event log in the combined list of log events of the block.
• L1 attributes deposited: keccak256(bytes32(uint256(1)), keccak256(l1BlockHash, bytes32(uint256(seqNumber)))). Where l1BlockHash refers to the L1 block hash of which the info attributes are deposited. And seqNumber = l2BlockNum - l2EpochStartBlockNum, where l2BlockNum is the L2 block number of the inclusion of the deposit tx in L2, and l2EpochStartBlockNum is the L2 block number of the first L2 block in the epoch.
• Upgrade-deposited: keccak256(bytes32(uint256(2)), keccak256(intent)). Where intent is a UTF-8 byte string, identifying the upgrade intent.

Without a sourceHash in a deposit, two different deposited transactions could have the same exact hash.

The outer keccak256 hashes the actual uniquely identifying information with a domain, to avoid collisions between different types of sources.

We do not use the sender's nonce to ensure uniqueness because this would require an extra L2 EVM state read from the execution engine during block-derivation.

Kinds of Deposited Transactions

Although we define only one new transaction type, we can distinguish between two kinds of deposited transactions, based on their positioning in the L2 block:

1. The first transaction MUST be a L1 attributes deposited transaction, followed by
2. an array of zero-or-more user-deposited transactions submitted to the deposit feed contract on L1 (called OptimismPortal). User-deposited transactions are only present in the first block of a L2 epoch.

We only define a single new transaction type in order to minimize modifications to L1 client software, and complexity in general.

Validation and Authorization of Deposited Transactions

As noted above, the deposited transaction type does not include a signature for validation. Rather, authorization is handled by the L2 chain derivation process, which when correctly applied will only derive transactions with a from address attested to by the logs of the L1 deposit contract.

Execution

In order to execute a deposited transaction:

First, the balance of the from account MUST be increased by the amount of mint. This is unconditional, and does not revert on deposit failure.

Then, the execution environment for a deposited transaction is initialized based on the transaction's attributes, in exactly the same manner as it would be for an EIP-155 transaction.

The deposit transaction is processed exactly like a type-2 (EIP-1559) transaction, with the exception of:

• No fee fields are verified: the deposit does not have any, as it pays for gas on L1.
• No nonce field is verified: the deposit does not have any, it's uniquely identified by its sourceHash.
• No access-list is processed: the deposit has no access-list, and it is thus processed as if the access-list is empty.
• No check if from is an Externally Owner Account (EOA): the deposit is ensured not to be an EOA through L1 address masking, this may change in future L1 contract-deployments to e.g. enable an account-abstraction like mechanism.
• Before the Regolith upgrade:
  • The execution output states a non-standard gas usage:
    • If isSystemTx is false: execution output states it uses gasLimit gas.
    • If isSystemTx is true: execution output states it uses 0 gas.
• No gas is refunded as ETH. (either by not refunding or utilizing the fact the gas-price of the deposit is 0)
• No transaction priority fee is charged. No payment is made to the block fee-recipient.
• No L1-cost fee is charged, as deposits are derived from L1 and do not have to be submitted as data back to it.
• No base fee is charged. The total base fee accounting does not change.

Note that this includes contract-deployment behavior like with regular transactions, and gas metering is the same (with the exception of fee related changes above), including metering of intrinsic gas.

Any non-EVM state-transition error emitted by the EVM execution is processed in a special way:

• It is transformed into an EVM-error: i.e. the deposit will always be included, but its receipt will indicate a failure if it runs into a non-EVM state-transition error, e.g. failure to transfer the specified value amount of ETH due to insufficient account-balance.
• The world state is rolled back to the start of the EVM processing, after the minting part of the deposit.
• The nonce of from in the world state is incremented by 1, making the error equivalent to a native EVM failure. Note that a previous nonce increment may have happened during EVM processing, but this would be rolled back first.

Finally, after the above processing, the execution post-processing runs the same: i.e. the gas pool and receipt are processed identical to a regular transaction. Starting with the Regolith upgrade however, the receipt of deposit transactions is extended with an additional depositNonce value, storing the nonce value of the from sender as registered before the EVM processing.

Note that the gas used as stated by the execution output is subtracted from the gas pool, but this execution output value has special edge cases before the Regolith upgrade.

Note for application developers: because CALLER and ORIGIN are set to from, the semantics of using the tx.origin == msg.sender check will not work to determine whether or not a caller is an EOA during a deposit transaction. Instead, the check could only be useful for identifying the first call in the L2 deposit transaction. However this check does still satisfy the common case in which developers are using this check to ensure that the CALLER is unable to execute code before and after the call.

Nonce Handling

Despite the lack of signature validation, we still increment the nonce of the from account when a deposit transaction is executed. In the context of a deposit-only roll up, this is not necessary for transaction ordering or replay prevention, however it maintains consistency with the use of nonces during contract creation. It may also simplify integration with downstream tooling (such as wallets and block explorers).

Deposit Receipt

Transaction receipts use standard typing as per EIP-2718. The Deposit transaction receipt type is equal to a regular receipt, but extended with an optional depositNonce field.

The RLP-encoded consensus-enforced fields are:

• postStateOrStatus (standard): this contains the transaction status, see EIP-658.
• cumulativeGasUsed (standard): gas used in the block thus far, including this transaction.
  • The actual gas used is derived from the difference in CumulativeGasUsed with the previous transaction.
  • Starting with Regolith, this accounts for the actual gas usage by the deposit, like regular transactions.
• bloom (standard): bloom filter of the transaction logs.
• logs (standard): log events emitted by the EVM processing.
• depositNonce (unique extension): Optional field. The deposit transaction persists the nonce used during execution.
• depositNonceVersion (unique extension): Optional field. The value must be 1 if the field is present
  • Before Canyon, these depositNonce & depositNonceVersion fields must always be omitted.
  • With Canyon, these depositNonce & depositNonceVersion fields must always be included.

Starting with Regolith, the receipt API responses utilize the receipt changes for more accurate response data:

• The depositNonce is included in the receipt JSON data in API responses
• For contract-deployments (when to == null), the depositNonce helps derive the correct contractAddress meta-data, instead of assuming the nonce was zero.
• The cumulativeGasUsed accounts for the actual gas usage, as metered in the EVM processing.

L1 Attributes Deposited Transaction

An L1 attributes deposited transaction is a deposit transaction sent to the L1 attributes predeployed contract.

This transaction MUST have the following values:

1. from is 0xdeaddeaddeaddeaddeaddeaddeaddeaddead0001 (the address of the L1 Attributes depositor account)
2. to is 0x4200000000000000000000000000000000000015 (the address of the L1 attributes predeployed contract).
3. mint is 0
4. value is 0
5. gasLimit is set to 150,000,000 prior to the Regolith upgrade, and 1,000,000 after.
6. isSystemTx is set to true prior to the Regolith upgrade, and false after.
7. data is an encoded call to the L1 attributes predeployed contract that depends on the upgrades that are active (see below).

This system-initiated transaction for L1 attributes is not charged any ETH for its allocated gasLimit, as it is considered part of state-transition processing.

L1 Attributes Deposited Transaction Calldata

L1 Attributes - Bedrock, Canyon, Delta

The data field of the L1 attributes deposited transaction is an ABI encoded call to the setL1BlockValues() function with correct values associated with the corresponding L1 block (cf. reference implementation).

Special Accounts on L2

The L1 attributes deposit transaction involves two special purpose accounts:

1. The L1 attributes depositor account
2. The L1 attributes predeployed contract

L1 Attributes Depositor Account

The depositor account is an EOA with no known private key. It has the address 0xdeaddeaddeaddeaddeaddeaddeaddeaddead0001. Its value is returned by the CALLER and ORIGIN opcodes during execution of the L1 attributes deposited transaction.

L1 Attributes Predeployed Contract

A predeployed contract on L2 at address 0x4200000000000000000000000000000000000015, which holds certain block variables from the corresponding L1 block in storage, so that they may be accessed during the execution of the subsequent deposited transactions.

The predeploy stores the following values:

• L1 block attributes:
  • number (uint64)
  • timestamp (uint64)
  • basefee (uint256)
  • hash (bytes32)
• sequenceNumber (uint64): This equals the L2 block number relative to the start of the epoch, i.e. the L2 block distance to the L2 block height that the L1 attributes last changed, and reset to 0 at the start of a new epoch.
• System configurables tied to the L1 block, see System configuration specification:
  • batcherHash (bytes32): A versioned commitment to the batch-submitter(s) currently operating.
  • overhead (uint256): The L1 fee overhead to apply to L1 cost computation of transactions in this L2 block.
  • scalar (uint256): The L1 fee scalar to apply to L1 cost computation of transactions in this L2 block.

The contract implements an authorization scheme, such that it only accepts state-changing calls from the depositor account.

The contract has the following solidity interface, and can be interacted with according to the contract ABI specification.

L1 Attributes Predeployed Contract: Reference Implementation

A reference implementation of the L1 Attributes predeploy contract can be found in L1Block.sol.

User-Deposited Transactions

User-deposited transactions are deposited transactions generated by the L2 Chain Derivation process. The content of each user-deposited transaction are determined by the corresponding TransactionDeposited event emitted by the deposit contract on L1.

1. from is unchanged from the emitted value (though it may have been transformed to an alias in OptimismPortal, the deposit feed contract).
2. to is any 20-byte address (including the zero address)
   • In case of a contract creation (cf. isCreation), this address is set to null.
3. mint is set to the emitted value.
4. value is set to the emitted value.
5. gaslimit is unchanged from the emitted value. It must be at least 21000.
6. isCreation is set to true if the transaction is a contract creation, false otherwise.
7. data is unchanged from the emitted value. Depending on the value of isCreation it is handled as either calldata or contract initialization code.
8. isSystemTx is set by the rollup node for certain transactions that have unmetered execution. It is false for user deposited transactions

Deposit Contract

The deposit contract is deployed to L1. Deposited transactions are derived from the values in the TransactionDeposited event(s) emitted by the deposit contract.

The deposit contract is responsible for maintaining the guaranteed gas market, charging deposits for gas to be used on L2, and ensuring that the total amount of guaranteed gas in a single L1 block does not exceed the L2 block gas limit.

The deposit contract handles two special cases:

1. A contract creation deposit, which is indicated by setting the isCreation flag to true. In the event that the to address is non-zero, the contract will revert.
2. A call from a contract account, in which case the from value is transformed to its L2 alias.

Address Aliasing

If the caller is a contract, the address will be transformed by adding 0x1111000000000000000000000000000000001111 to it. The math is unchecked and done on a Solidity uint160 so the value will overflow. This prevents attacks in which a contract on L1 has the same address as a contract on L2 but doesn't have the same code. We can safely ignore this for EOAs because they're guaranteed to have the same "code" (i.e. no code at all). This also makes it possible for users to interact with contracts on L2 even when the Sequencer is down.

Deposit Contract Implementation: Optimism Portal

A reference implementation of the deposit contract can be found in OptimismPortal.sol.

Withdrawals

Table of Contents

• Overview
• Withdrawal Flow
  • On L2
  • On L1
• The L2ToL1MessagePasser Contract
  • Addresses are not Aliased on Withdrawals
• The Optimism Portal Contract
• Withdrawal Verification and Finalization
• Security Considerations
  • Key Properties of Withdrawal Verification
  • Handling Successfully Verified Messages That Fail When Relayed
  • OptimismPortal can send arbitrary messages on L1

Overview

Withdrawals are cross domain transactions which are initiated on L2, and finalized by a transaction executed on L1. Notably, withdrawals may be used by an L2 account to call an L1 contract, or to transfer ETH from an L2 account to an L1 account.

Vocabulary note: withdrawal can refer to the transaction at various stages of the process, but we introduce more specific terms to differentiate:

• A withdrawal initiating transaction refers specifically to a transaction on L2 sent to the Withdrawals predeploy.
• A withdrawal proving transaction refers specifically to an L1 transaction which proves the withdrawal is correct (that it has been included in a merkle tree whose root is available on L1).
• A withdrawal finalizing transaction refers specifically to an L1 transaction which finalizes and relays the withdrawal.

Withdrawals are initiated on L2 via a call to the Message Passer predeploy contract, which records the important properties of the message in its storage. Withdrawals are proven on L1 via a call to the OptimismPortal, which proves the inclusion of this withdrawal message. Withdrawals are finalized on L1 via a call to the OptimismPortal contract, which verifies that the fault challenge period has passed since the withdrawal message has been proved.

In this way, withdrawals are different from deposits which make use of a special transaction type in the execution engine client. Rather, withdrawals transaction must use smart contracts on L1 for finalization.

Withdrawal Flow

We first describe the end to end flow of initiating and finalizing a withdrawal:

On L2

An L2 account sends a withdrawal message (and possibly also ETH) to the L2ToL1MessagePasser predeploy contract. This is a very simple contract that stores the hash of the withdrawal data.

On L1

1. A relayer submits a withdrawal proving transaction with the required inputs to the OptimismPortal contract. The relayer is not necessarily the same entity which initiated the withdrawal on L2. These inputs include the withdrawal transaction data, inclusion proofs, and a block number. The block number must be one for which an L2 output root exists, which commits to the withdrawal as registered on L2.
2. The OptimismPortal contract retrieves the output root for the given block number from the L2OutputOracle's getL2Output() function, and performs the remainder of the verification process internally.
3. If proof verification fails, the call reverts. Otherwise the hash is recorded to prevent it from being re-proven. Note that the withdrawal can be proven more than once if the corresponding output root changes.
4. After the withdrawal is proven, it enters a 7 day challenge period, allowing time for other network participants to challenge the integrity of the corresponding output root.
5. Once the challenge period has passed, a relayer submits a withdrawal finalizing transaction to the OptimismPortal contract. The relayer doesn't need to be the same entity that initiated the withdrawal on L2.
6. The OptimismPortal contract receives the withdrawal transaction data and verifies that the withdrawal has both been proven and passed the challenge period.
7. If the requirements are not met, the call reverts. Otherwise the call is forwarded, and the hash is recorded to prevent it from being replayed.

The L2ToL1MessagePasser Contract

A withdrawal is initiated by calling the L2ToL1MessagePasser contract's initiateWithdrawal function. The L2ToL1MessagePasser is a simple predeploy contract at 0x4200000000000000000000000000000000000016 which stores messages to be withdrawn.

interface L2ToL1MessagePasser {
    event MessagePassed(
        uint256 indexed nonce, // this is a global nonce value for all withdrawal messages
        address indexed sender,
        address indexed target,
        uint256 value,
        uint256 gasLimit,
        bytes data,
        bytes32 withdrawalHash
    );

    event WithdrawerBalanceBurnt(uint256 indexed amount);

    function burn() external;

    function initiateWithdrawal(address _target, uint256 _gasLimit, bytes memory _data) payable external;

    function messageNonce() public view returns (uint256);

    function sentMessages(bytes32) view external returns (bool);
}

The MessagePassed event includes all of the data that is hashed and stored in the sentMessages mapping, as well as the hash itself.

Addresses are not Aliased on Withdrawals

When a contract makes a deposit, the sender's address is aliased. The same is not true of withdrawals, which do not modify the sender's address. The difference is that:

• on L2, the deposit sender's address is returned by the CALLER opcode, meaning a contract cannot easily tell if the call originated on L1 or L2, whereas
• on L1, the withdrawal sender's address is accessed by calling the l2Sender() function on the OptimismPortal contract.

Calling l2Sender() removes any ambiguity about which domain the call originated from. Still, developers will need to recognize that having the same address does not imply that a contract on L2 will behave the same as a contract on L1.

The Optimism Portal Contract

The Optimism Portal serves as both the entry and exit point to the Optimism L2. It is a contract which inherits from the OptimismPortal contract, and in addition provides the following interface for withdrawals:

• WithdrawalTransaction type
• OutputRootProof type

interface OptimismPortal {

    event WithdrawalFinalized(bytes32 indexed withdrawalHash, bool success);


    function l2Sender() returns(address) external;

    function proveWithdrawalTransaction(
        Types.WithdrawalTransaction memory _tx,
        uint256 _l2OutputIndex,
        Types.OutputRootProof calldata _outputRootProof,
        bytes[] calldata _withdrawalProof
    ) external;

    function finalizeWithdrawalTransaction(
        Types.WithdrawalTransaction memory _tx
    ) external;
}

Withdrawal Verification and Finalization

The following inputs are required to prove and finalize a withdrawal:

• Withdrawal transaction data:
  • nonce: Nonce for the provided message.
  • sender: Message sender address on L2.
  • target: Target address on L1.
  • value: ETH to send to the target.
  • data: Data to send to the target.
  • gasLimit: Gas to be forwarded to the target.
• Proof and verification data:
  • l2OutputIndex: The index in the L2 outputs where the applicable output root may be found.
  • outputRootProof: Four bytes32 values which are used to derive the output root.
  • withdrawalProof: An inclusion proof for the given withdrawal in the L2ToL1MessagePasser contract.

These inputs must satisfy the following conditions:

1. The l2OutputIndex must be the index in the L2 outputs that contains the applicable output root.
2. L2OutputOracle.getL2Output(l2OutputIndex) returns a non-zero OutputProposal.
3. The keccak256 hash of the outputRootProof values is equal to the outputRoot.
4. The withdrawalProof is a valid inclusion proof demonstrating that a hash of the Withdrawal transaction data is contained in the storage of the L2ToL1MessagePasser contract on L2.

Security Considerations

Key Properties of Withdrawal Verification

1. It should not be possible to 'double spend' a withdrawal, ie. to relay a withdrawal on L1 which does not correspond to a message initiated on L2. For reference, see this writeup of a vulnerability of this type found on Polygon.

2. For each withdrawal initiated on L2 (i.e. with a unique messageNonce()), the following properties must hold:

   1. It should only be possible to prove the withdrawal once, unless the outputRoot for the withdrawal has changed.
   2. It should only be possible to finalize the withdrawal once.
   3. It should not be possible to relay the message with any of its fields modified, ie.
      1. Modifying the sender field would enable a 'spoofing' attack.
      2. Modifying the target, data, or value fields would enable an attacker to dangerously change the intended outcome of the withdrawal.
      3. Modifying the gasLimit could make the cost of relaying too high, or allow the relayer to cause execution to fail (out of gas) in the target.

Handling Successfully Verified Messages That Fail When Relayed

If the execution of the relayed call fails in the target contract, it is unfortunately not possible to determine whether or not it was 'supposed' to fail, and whether or not it should be 'replayable'. For this reason, and to minimize complexity, we have not provided any replay functionality, this may be implemented in external utility contracts if desired.

OptimismPortal can send arbitrary messages on L1

The L2ToL1MessagePasser contract's initiateWithdrawal function accepts a _target address and _data bytes, which is passed to a CALL opcode on L1 when finalizeWithdrawalTransaction is called after the challenge period. This means that, by design, the OptimismPortal contract can be used to send arbitrary transactions on the L1, with the OptimismPortal as the msg.sender.

This means users of the OptimismPortal contract should be careful what permissions they grant to the portal. For example, any ERC20 tokens mistakenly sent to the OptimismPortal contract are essentially lost, as they can be claimed by anybody that pre-approves transfers of this token out of the portal, using the L2 to initiate the approval and the L1 to prove and finalize the approval (after the challenge period).

Guaranteed Gas Fee Market

Table of Contents

• Overview
• Gas Stipend
• Default Values
• Limiting Guaranteed Gas
• Rationale for burning L1 Gas
• On Preventing Griefing Attacks

Overview

Deposited transactions are transactions on L2 that are initiated on L1. The gas that they use on L2 is bought on L1 via a gas burn (or a direct payment in the future). We maintain a fee market and hard cap on the amount of gas provided to all deposits in a single L1 block.

The gas provided to deposited transactions is sometimes called "guaranteed gas". The gas provided to deposited transactions is unique in the regard that it is not refundable. It cannot be refunded as it is sometimes paid for with a gas burn and there may not be any ETH left to refund.

The guaranteed gas is composed of a gas stipend, and of any guaranteed gas the user would like to purchase (on L1) on top of that.

Guaranteed gas on L2 is bought in the following manner. An L2 gas price is calculated via an EIP-1559-style algorithm. The total amount of ETH required to buy that gas is then calculated as (guaranteed gas * L2 deposit base fee). The contract then accepts that amount of ETH (in a future upgrade) or (only method right now), burns an amount of L1 gas that corresponds to the L2 cost (L2 cost / L1 base fee). The L2 gas price for guaranteed gas is not synchronized with the base fee on L2 and will likely be different.

Gas Stipend

To offset the gas spent on the deposit event, we credit gas spent * L1 base fee ETH to the cost of the L2 gas, where gas spent is the amount of L1 gas spent processing the deposit. If the ETH value of this credit is greater than the ETH value of the requested guaranteed gas (requested guaranteed gas * L2 gas price), no L1 gas is burnt.

Default Values

Limiting Guaranteed Gas

The total amount of guaranteed gas that can be bought in a single L1 block must be limited to prevent a denial of service attack against L2 as well as ensure the total amount of guaranteed gas stays below the L2 block gas limit.

We set a guaranteed gas limit of MAX_RESOURCE_LIMIT gas per L1 block and a target of MAX_RESOURCE_LIMIT / ELASTICITY_MULTIPLIER gas per L1 block. These numbers enabled occasional large transactions while staying within our target and maximum gas usage on L2.

Because the amount of guaranteed L2 gas that can be purchased in a single block is now limited, we implement an EIP-1559-style fee market to reduce congestion on deposits. By setting the limit at a multiple of the target, we enable deposits to temporarily use more L2 gas at a greater cost.

# Pseudocode to update the L2 deposit base fee and cap the amount of guaranteed gas
# bought in a block. Calling code must handle the gas burn and validity checks on
# the ability of the account to afford this gas.

# prev_base fee is a u128, prev_bought_gas and prev_num are u64s
prev_base_fee, prev_bought_gas, prev_num = <values from previous update>
now_num = block.number

# Clamp the full base fee to a specific range. The minimum value in the range should be around 100-1000
# to enable faster responses in the base fee. This replaces the `max` mechanism in the ethereum 1559
# implementation (it also serves to enable the base fee to increase if it is very small).
def clamp(v: i256, min: u128, max: u128) -> u128:
    if v < i256(min):
        return min
    elif v > i256(max):
        return max
    else:
        return u128(v)

# If this is a new block, update the base fee and reset the total gas
# If not, just update the total gas
if prev_num == now_num:
    now_base_fee = prev_base_fee
    now_bought_gas = prev_bought_gas + requested_gas
elif prev_num != now_num:
    # Width extension and conversion to signed integer math
    gas_used_delta = int128(prev_bought_gas) - int128(TARGET_RESOURCE_LIMIT)
    # Use truncating (round to 0) division - solidity's default.
    # Sign extend gas_used_delta & prev_base_fee to 256 bits to avoid overflows here.
    base_fee_per_gas_delta = prev_base_fee * gas_used_delta / TARGET_RESOURCE_LIMIT / BASE_FEE_MAX_CHANGE_DENOMINATOR
    now_base_fee_wide = prev_base_fee + base_fee_per_gas_delta

    now_base_fee = clamp(now_base_fee_wide, min=MINIMUM_BASE_FEE, max=UINT_128_MAX_VALUE)
    now_bought_gas =  requested_gas

    # If we skipped multiple blocks between the previous block and now update the base fee again.
    # This is not exactly the same as iterating the above function, but quite close for reasonable
    # gas target values. It is also constant time wrt the number of missed blocks which is important
    # for keeping gas usage stable.
    if prev_num + 1 < now_num:
        n = now_num - prev_num - 1
        # Apply 7/8 reduction to prev_base_fee for the n empty blocks in a row.
        now_base_fee_wide = now_base_fee * pow(1-(1/BASE_FEE_MAX_CHANGE_DENOMINATOR), n)
        now_base_fee = clamp(now_base_fee_wide, min=MINIMUM_BASE_FEE, max=type(uint128).max)

require(now_bought_gas < MAX_RESOURCE_LIMIT)

store_values(now_base_fee, now_bought_gas, now_num)

Rationale for burning L1 Gas

There must be a sybil resistance mechanism for usage of the network. If it is very cheap to get guaranteed gas on L2, then it would be possible to spam the network. Burning a dynamic amount of gas on L1 acts as a sybil resistance mechanism as it becomes more expensive with more demand.

If we collect ETH directly to pay for L2 gas, every (indirect) caller of the deposit function will need to be marked with the payable selector. This won't be possible for many existing projects. Unfortunately this is quite wasteful. As such, we will provide two options to buy L2 gas:

1. Burn L1 Gas
2. Send ETH to the Optimism Portal (Not yet supported)

The payable version (Option 2) will likely have discount applied to it (or conversely, #1 has a premium applied to it).

For the initial release of bedrock, only #1 is supported.

On Preventing Griefing Attacks

The cost of purchasing all of the deposit gas in every block must be expensive enough to prevent attackers from griefing all deposits to the network. An attacker would observe a deposit in the mempool and frontrun it with a deposit that purchases enough gas such that the other deposit reverts. The smaller the max resource limit is, the easier this attack is to pull off. This attack is mitigated by having a large resource limit as well as a large elasticity multiplier. This means that the target resource usage is kept small, giving a lot of room for the deposit base fee to rise when the max resource limit is being purchased.

This attack should be too expensive to pull off in practice, but if an extremely wealthy adversary does decide to grief network deposits for an extended period of time, efforts will be placed to ensure that deposits are able to be processed on the network.

L2 Output Root Proposals Specification

Table of Contents

• Overview
• Proposing L2 Output Commitments
  • L2OutputOracle v1.0.0
• L2 Output Commitment Construction
• L2 Output Oracle Smart Contract
  • Configuration
• Security Considerations
  • L1 Reorgs

Overview

After processing one or more blocks the outputs will need to be synchronized with the settlement layer (L1) for trustless execution of L2-to-L1 messaging, such as withdrawals. These output proposals act as the bridge's view into the L2 state. Actors called "Proposers" submit the output roots to the settlement layer (L1) and can be contested with a fault proof, with a bond at stake if the proof is wrong. The op-proposer in one such implementation of a proposer.

Note: Fault proofs on Optimism are not fully specified at this time. Although fault proof construction and verification is implemented in Cannon, the fault proof game specification and integration of a output-root challenger into the rollup-node are part of later specification milestones.

Proposing L2 Output Commitments

The proposer's role is to construct and submit output roots, which are commitments to the L2's state, to the L2OutputOracle contract on L1 (the settlement layer). To do this, the proposer periodically queries the rollup node for the latest output root derived from the latest finalized L1 block. It then takes the output root and submits it to the L2OutputOracle contract on the settlement layer (L1).

L2OutputOracle v1.0.0

The submission of output proposals is permissioned to a single account. It is expected that this account will continue to submit output proposals over time to ensure that user withdrawals do not halt.

The L2 output proposer is expected to submit output roots on a deterministic interval based on the configured SUBMISSION_INTERVAL in the L2OutputOracle. The larger the SUBMISSION_INTERVAL, the less often L1 transactions need to be sent to the L2OutputOracle contract, but L2 users will need to wait a bit longer for an output root to be included in L1 (the settlement layer) that includes their intention to withdraw from the system.

The honest op-proposer algorithm assumes a connection to the L2OutputOracle contract to know the L2 block number that corresponds to the next output proposal that must be submitted. It also assumes a connection to an op-node to be able to query the optimism_syncStatus RPC endpoint.

import time

while True:
    next_checkpoint_block = L2OutputOracle.nextBlockNumber()
    rollup_status = op_node_client.sync_status()
    if rollup_status.finalized_l2.number >= next_checkpoint_block:
        output = op_node_client.output_at_block(next_checkpoint_block)
        tx = send_transaction(output)
    time.sleep(poll_interval)

A CHALLENGER account can delete multiple output roots by calling the deleteL2Outputs() function and specifying the index of the first output to delete, this will also delete all subsequent outputs.

L2 Output Commitment Construction

The output_root is a 32 byte string, which is derived based on the a versioned scheme:

output_root = keccak256(version_byte || payload)

where:

1. version_byte (bytes32) a simple version string which increments anytime the construction of the output root is changed.

2. payload (bytes) is a byte string of arbitrary length.

In the initial version of the output commitment construction, the version is bytes32(0), and the payload is defined as:

payload = state_root || withdrawal_storage_root || latest_block_hash

where:

1. The latest_block_hash (bytes32) is the block hash for the latest L2 block.

2. The state_root (bytes32) is the Merkle-Patricia-Trie (MPT) root of all execution-layer accounts. This value is frequently used and thus elevated closer to the L2 output root, which removes the need to prove its inclusion in the pre-image of the latest_block_hash. This reduces the merkle proof depth and cost of accessing the L2 state root on L1.

3. The withdrawal_storage_root (bytes32) elevates the Merkle-Patricia-Trie (MPT) root of the Message Passer contract storage. Instead of making an MPT proof for a withdrawal against the state root (proving first the storage root of the L2toL1MessagePasser against the state root, then the withdrawal against that storage root), we can prove against the L2toL1MessagePasser's storage root directly, thus reducing the verification cost of withdrawals on L1.

   After Isthmus hard fork, the withdrawal_storage_root is present in the block header as withdrawalsRoot and can be used directly, instead of computing the storage root of the L2toL1MessagePasser contract.

   Similarly, if Isthmus hard fork is active at the genesis block, the withdrawal_storage_root is present in the block header as withdrawalsRoot.

L2 Output Oracle Smart Contract

L2 blocks are produced at a constant rate of L2_BLOCK_TIME (2 seconds). A new L2 output MUST be appended to the chain once per SUBMISSION_INTERVAL which is based on a number of blocks. The exact number is yet to be determined, and will depend on the design of the fault proving game.

The L2 Output Oracle contract implements the following interface:

/**
 * @notice The number of the first L2 block recorded in this contract.
 */
uint256 public startingBlockNumber;

/**
 * @notice The timestamp of the first L2 block recorded in this contract.
 */
uint256 public startingTimestamp;

/**
 * @notice Accepts an L2 outputRoot and the timestamp of the corresponding L2 block. The
 * timestamp must be equal to the current value returned by `nextTimestamp()` in order to be
 * accepted.
 * This function may only be called by the Proposer.
 *
 * @param _l2Output      The L2 output of the checkpoint block.
 * @param _l2BlockNumber The L2 block number that resulted in _l2Output.
 * @param _l1Blockhash   A block hash which must be included in the current chain.
 * @param _l1BlockNumber The block number with the specified block hash.
*/
  function proposeL2Output(
      bytes32 _l2Output,
      uint256 _l2BlockNumber,
      bytes32 _l1Blockhash,
      uint256 _l1BlockNumber
  )

/**
 * @notice Deletes all output proposals after and including the proposal that corresponds to
 *         the given output index. Only the challenger address can delete outputs.
 *
 * @param _l2OutputIndex Index of the first L2 output to be deleted. All outputs after this
 *                       output will also be deleted.
 */
function deleteL2Outputs(uint256 _l2OutputIndex) external

/**
 * @notice Computes the block number of the next L2 block that needs to be checkpointed.
 */
function nextBlockNumber() public view returns (uint256)

Configuration

The startingBlockNumber must be at least the number of the first Bedrock block. The startingTimestamp MUST be the same as the timestamp of the start block.

The first outputRoot proposed will thus be at height startingBlockNumber + SUBMISSION_INTERVAL

Security Considerations

L1 Reorgs

If the L1 has a reorg after an output has been generated and submitted, the L2 state and correct output may change leading to a faulty proposal. This is mitigated against by allowing the proposer to submit an L1 block number and hash to the Output Oracle when appending a new output; in the event of a reorg, the block hash will not match that of the block with that number and the call will revert.

L2 Execution Engine

Table of Contents

• 1559 Parameters
• Deposited transaction processing
  • Deposited transaction boundaries
• Fees
  • Fee Vaults
  • Priority fees (Sequencer Fee Vault)
  • Base fees (Base Fee Vault)
  • L1-Cost fees (L1 Fee Vault)
    • Pre-Ecotone
    • Ecotone L1-Cost fee changes (EIP-4844 DA)
• Engine API
  • engine_forkchoiceUpdatedV2
    • Extended PayloadAttributesV2
  • engine_forkchoiceUpdatedV3
    • Extended PayloadAttributesV3
  • engine_newPayloadV2
  • engine_newPayloadV3
  • engine_getPayloadV2
  • engine_getPayloadV3
    • Extended Response
  • engine_signalSuperchainV1
• Networking
• Sync
  • Happy-path sync
  • Worst-case sync
• Ecotone: disable Blob-transactions
• Ecotone: Beacon Block Root
• P2P Modifications

This document outlines the modifications, configuration and usage of a L1 execution engine for L2.

1559 Parameters

The execution engine must be able to take a per chain configuration which specifies the EIP-1559 Denominator and EIP-1559 elasticity. After Canyon it should also take a new value EIP1559DenominatorCanyon and use that as the denominator in the 1559 formula rather than the prior denominator.

The formula for EIP-1559 is not otherwise modified.

Deposited transaction processing

The Engine interfaces abstract away transaction types with EIP-2718.

To support rollup functionality, processing of a new Deposit TransactionType is implemented by the engine, see the deposits specification.

This type of transaction can mint L2 ETH, run EVM, and introduce L1 information to enshrined contracts in the execution state.

Deposited transaction boundaries

Transactions cannot be blindly trusted, trust is established through authentication. Unlike other transaction types deposits are not authenticated by a signature: the rollup node authenticates them, outside of the engine.

To process deposited transactions safely, the deposits MUST be authenticated first:

• Ingest directly through trusted Engine API
• Part of sync towards a trusted block hash (trusted through previous Engine API instruction)

Deposited transactions MUST never be consumed from the transaction pool. The transaction pool can be disabled in a deposits-only rollup

Fees

Sequenced transactions (i.e. not applicable to deposits) are charged with 3 types of fees: priority fees, base fees, and L1-cost fees.

Fee Vaults

The three types of fees are collected in 3 distinct L2 fee-vault deployments for accounting purposes: fee payments are not registered as internal EVM calls, and thus distinguished better this way.

These are hardcoded addresses, pointing at pre-deployed proxy contracts. The proxies are backed by vault contract deployments, based on FeeVault, to route vault funds to L1 securely.

Priority fees (Sequencer Fee Vault)

Priority fees follow the eip-1559 specification, and are collected by the fee-recipient of the L2 block. The block fee-recipient (a.k.a. coinbase address) is set to the Sequencer Fee Vault address.

Base fees (Base Fee Vault)

Base fees largely follow the eip-1559 specification, with the exception that base fees are not burned, but add up to the Base Fee Vault ETH account balance.

L1-Cost fees (L1 Fee Vault)

The protocol funds batch-submission of sequenced L2 transactions by charging L2 users an additional fee based on the estimated batch-submission costs. This fee is charged from the L2 transaction-sender ETH balance, and collected into the L1 Fee Vault.

The exact L1 cost function to determine the L1-cost fee component of a L2 transaction depends on the upgrades that are active.

Pre-Ecotone

Before Ecotone activation, L1 cost is calculated as: (rollupDataGas + l1FeeOverhead) * l1BaseFee * l1FeeScalar / 1e6 (big-int computation, result in Wei and uint256 range) Where:

• rollupDataGas is determined from the full encoded transaction (standard EIP-2718 transaction encoding, including signature fields):
  • Before Regolith fork: rollupDataGas = zeroes * 4 + (ones + 68) * 16
    • The addition of 68 non-zero bytes is a remnant of a pre-Bedrock L1-cost accounting function, which accounted for the worst-case non-zero bytes addition to complement unsigned transactions, unlike Bedrock.
  • With Regolith fork: rollupDataGas = zeroes * 4 + ones * 16
• l1FeeOverhead is the Gas Price Oracle overhead value.
• l1FeeScalar is the Gas Price Oracle scalar value.
• l1BaseFee is the L1 base fee of the latest L1 origin registered in the L2 chain.

Note that the rollupDataGas uses the same byte cost accounting as defined in eip-2028, except the full L2 transaction now counts towards the bytes charged in the L1 calldata. This behavior matches pre-Bedrock L1-cost estimation of L2 transactions.

Compression, batching, and intrinsic gas costs of the batch transactions are accounted for by the protocol with the Gas Price Oracle overhead and scalar parameters.

The Gas Price Oracle l1FeeOverhead and l1FeeScalar, as well as the l1BaseFee of the L1 origin, can be accessed in two interchangeable ways:

• read from the deposited L1 attributes (l1FeeOverhead, l1FeeScalar, basefee) of the current L2 block
• read from the L1 Block Info contract (0x4200000000000000000000000000000000000015)
  • using the respective solidity uint256-getter functions (l1FeeOverhead, l1FeeScalar, basefee)
  • using direct storage-reads:
    • L1 basefee as big-endian uint256 in slot 1
    • Overhead as big-endian uint256 in slot 5
    • Scalar as big-endian uint256 in slot 6

Ecotone L1-Cost fee changes (EIP-4844 DA)

Ecotone allows posting batches via Blobs which are subject to a new fee market. To account for this feature, L1 cost is computed as:

(zeroes*4 + ones*16) * (16*l1BaseFee*l1BaseFeeScalar + l1BlobBaseFee*l1BlobBaseFeeScalar) / 16e6

Where:

• the computation is an unlimited precision integer computation, with the result in Wei and having uint256 range.

• zeroes and ones are the count of zero and non-zero bytes respectively in the full encoded signed transaction.

• l1BaseFee is the L1 base fee of the latest L1 origin registered in the L2 chain.

• l1BlobBaseFee is the blob gas price, computed as described in EIP-4844 from the header of the latest registered L1 origin block.

Conceptually what the above function captures is the formula below, where compressedTxSize = (zeroes*4 + ones*16) / 16 can be thought of as a rough approximation of how many bytes the transaction occupies in a compressed batch.

(compressedTxSize) * (16*l1BaseFee*lBaseFeeScalar + l1BlobBaseFee*l1BlobBaseFeeScalar) / 1e6

The precise cost function used by Ecotone at the top of this section preserves precision under integer arithmetic by postponing the inner division by 16 until the very end.

The two base fee values and their respective scalars can be accessed in two interchangeable ways:

• read from the deposited L1 attributes (l1BaseFeeScalar, l1BlobBaseFeeScalar, basefee, blobBaseFee) of the current L2 block
• read from the L1 Block Info contract (0x4200000000000000000000000000000000000015)
  • using the respective solidity getter functions
  • using direct storage-reads:
    • basefee uint256 in slot 1
    • blobBaseFee uint256 in slot 7
    • l1BaseFeeScalar big-endian uint32 slot 3 at offset 12
    • l1BlobBaseFeeScalar big-endian uint32 in slot 3 at offset 8

Engine API

engine_forkchoiceUpdatedV2

This updates which L2 blocks the engine considers to be canonical (forkchoiceState argument), and optionally initiates block production (payloadAttributes argument).

Within the rollup, the types of forkchoice updates translate as:

• headBlockHash: block hash of the head of the canonical chain. Labeled "unsafe" in user JSON-RPC. Nodes may apply L2 blocks out of band ahead of time, and then reorg when L1 data conflicts.
• safeBlockHash: block hash of the canonical chain, derived from L1 data, unlikely to reorg.
• finalizedBlockHash: irreversible block hash, matches lower boundary of the dispute period.

To support rollup functionality, one backwards-compatible change is introduced to engine_forkchoiceUpdatedV2: the extended PayloadAttributesV2

Extended PayloadAttributesV2

PayloadAttributesV2 is extended to:

PayloadAttributesV2: {
    timestamp: QUANTITY
    random: DATA (32 bytes)
    suggestedFeeRecipient: DATA (20 bytes)
    withdrawals: array of WithdrawalV1
    transactions: array of DATA
    noTxPool: bool
    gasLimit: QUANTITY or null
}

The type notation used here refers to the HEX value encoding used by the Ethereum JSON-RPC API specification, as this structure will need to be sent over JSON-RPC. array refers to a JSON array.

Each item of the transactions array is a byte list encoding a transaction: TransactionType || TransactionPayload or LegacyTransaction, as defined in EIP-2718. This is equivalent to the transactions field in ExecutionPayloadV2

The transactions field is optional:

• If empty or missing: no changes to engine behavior. The sequencers will (if enabled) build a block by consuming transactions from the transaction pool.
• If present and non-empty: the payload MUST be produced starting with this exact list of transactions. The rollup driver determines the transaction list based on deterministic L1 inputs.

The noTxPool is optional as well, and extends the transactions meaning:

• If false, the execution engine is free to pack additional transactions from external sources like the tx pool into the payload, after any of the transactions. This is the default behavior a L1 node implements.
• If true, the execution engine must not change anything about the given list of transactions.

If the transactions field is present, the engine must execute the transactions in order and return STATUS_INVALID if there is an error processing the transactions. It must return STATUS_VALID if all of the transactions could be executed without error. Note: The state transition rules have been modified such that deposits will never fail so if engine_forkchoiceUpdatedV2 returns STATUS_INVALID it is because a batched transaction is invalid.

The gasLimit is optional w.r.t. compatibility with L1, but required when used as rollup. This field overrides the gas limit used during block-building. If not specified as rollup, a STATUS_INVALID is returned.

engine_forkchoiceUpdatedV3

See engine_forkchoiceUpdatedV2 for a description of the forkchoice updated method. engine_forkchoiceUpdatedV3 must only be called with Ecotone payload.

To support rollup functionality, one backwards-compatible change is introduced to engine_forkchoiceUpdatedV3: the extended PayloadAttributesV3

Extended PayloadAttributesV3

PayloadAttributesV3 is extended to:

PayloadAttributesV3: {
    timestamp: QUANTITY
    random: DATA (32 bytes)
    suggestedFeeRecipient: DATA (20 bytes)
    withdrawals: array of WithdrawalV1
    parentBeaconBlockRoot: DATA (32 bytes)
    transactions: array of DATA
    noTxPool: bool
    gasLimit: QUANTITY or null
}

The requirements of this object are the same as extended PayloadAttributesV2 with the addition of parentBeaconBlockRoot which is the parent beacon block root from the L1 origin block of the L2 block.

Starting at Ecotone, the parentBeaconBlockRoot must be set to the L1 origin parentBeaconBlockRoot, or a zero bytes32 if the Dencun functionality with parentBeaconBlockRoot is not active on L1.

engine_newPayloadV2

No modifications to engine_newPayloadV2. Applies a L2 block to the engine state.

engine_newPayloadV3

engine_newPayloadV3 applies an Ecotone L2 block to the engine state. There are no modifications to this API. engine_newPayloadV3 must only be called with Ecotone payload.

The additional parameters should be set as follows:

• expectedBlobVersionedHashes MUST be an empty array.
• parentBeaconBlockRoot MUST be the parent beacon block root from the L1 origin block of the L2 block.

engine_getPayloadV2

No modifications to engine_getPayloadV2. Retrieves a payload by ID, prepared by engine_forkchoiceUpdatedV2 when called with payloadAttributes.

engine_getPayloadV3

engine_getPayloadV3 retrieves a payload by ID, prepared by engine_forkchoiceUpdatedV3 when called with payloadAttributes. engine_getPayloadV3 must only be called with Ecotone payload.

Extended Response

The response is extended to:

{
    executionPayload: ExecutionPayload
    blockValue: QUANTITY
    blobsBundle: BlobsBundle
    shouldOverrideBuilder: BOOLEAN
    parentBeaconBlockRoot: DATA (32 bytes)
}

In Ecotone it MUST be set to the parentBeaconBlockRoot from the L1 Origin block of the L2 block.

engine_signalSuperchainV1

Optional extension to the Engine API. Signals superchain information to the Engine: V1 signals which protocol version is recommended and required.

Types:

SuperchainSignal: {
  recommended: ProtocolVersion;
  required: ProtocolVersion;
}

ProtocolVersion: encoded for RPC as defined in Protocol Version format specification.

Parameters:

• signal: SuperchainSignal, the signaled superchain information.

Returns:

• ProtocolVersion: the latest supported OP-Stack protocol version of the execution engine.

The execution engine SHOULD warn the user when the recommended version is newer than the current version supported by the execution engine.

The execution engine SHOULD take safety precautions if it does not meet the required protocol version. This may include halting the engine, with consent of the execution engine operator.

Networking

The execution engine can acquire all data through the rollup node, as derived from L1: P2P networking is strictly optional.

However, to not bottleneck on L1 data retrieval speed, the P2P network functionality SHOULD be enabled, serving:

• Peer discovery (Disc v5)
• eth/66:
  • Transaction pool (consumed by sequencer nodes)
  • State sync (happy-path for fast trustless db replication)
  • Historical block header and body retrieval
  • New blocks are acquired through the consensus layer instead (rollup node)

No modifications to L1 network functionality are required, except configuration:

• networkID: Distinguishes the L2 network from L1 and testnets. Equal to the chainID of the rollup network.
• Activate Merge fork: Enables Engine API and disables propagation of blocks, as block headers cannot be authenticated without consensus layer.
• Bootnode list: DiscV5 is a shared network, bootstrap is faster through connecting with L2 nodes first.

Sync

The execution engine can operate sync in different ways:

• Happy-path: rollup node informs engine of the desired chain head as determined by L1, completes through engine P2P.
• Worst-case: rollup node detects stalled engine, completes sync purely from L1 data, no peers required.

The happy-path is more suitable to bring new nodes online quickly, as the engine implementation can sync state faster through methods like snap-sync.

Happy-path sync

1. The rollup node informs the engine of the L2 chain head, unconditionally (part of regular node operation):
   • Bedrock / Canyon / Delta Payloads
     • engine_newPayloadV2 is called with latest L2 block received from P2P.
     • engine_forkchoiceUpdatedV2 is called with the current unsafe/safe/finalized L2 block hashes.
   • Ecotone Payloads
     • engine_newPayloadV3 is called with latest L2 block received from P2P.
     • engine_forkchoiceUpdatedV3 is called with the current unsafe/safe/finalized L2 block hashes.
2. The engine requests headers from peers, in reverse till the parent hash matches the local chain
3. The engine catches up: a) A form of state sync is activated towards the finalized or head block hash b) A form of block sync pulls block bodies and processes towards head block hash

The exact P2P based sync is out of scope for the L2 specification: the operation within the engine is the exact same as with L1 (although with an EVM that supports deposits).

Worst-case sync

1. Engine is out of sync, not peered and/or stalled due other reasons.
2. The rollup node maintains latest head from engine (poll eth_getBlockByNumber and/or maintain a header subscription)
3. The rollup node activates sync if the engine is out of sync but not syncing through P2P (eth_syncing)
4. The rollup node inserts blocks, derived from L1, one by one, potentially adapting to L1 reorg(s), as outlined in the rollup node spec.

Ecotone: disable Blob-transactions

EIP-4844 introduces Blob transactions: featuring all the functionality of an EIP-1559 transaction, plus a list of "blobs": "Binary Large Object", i.e. a dedicated data type for serving Data-Availability as base-layer.

With the Ecotone upgrade, all Cancun L1 execution features are enabled, with EIP-4844 as exception: as a L2, the OP-Stack does not serve blobs, and thus disables this new transaction type.

EIP-4844 is disabled as following:

• Transaction network-layer announcements, announcing blob-type transactions, are ignored.
• Transactions of the blob-type, through the RPC or otherwise, are not allowed into the transaction pool.
• Block-building code does not select EIP-4844 transactions.
• An L2 block state-transition with EIP-4844 transactions is invalid.

The BLOBBASEFEE opcode is present but its semantics are altered because there are no blobs processed by L2. The opcode will always push a value of 1 onto the stack.

Ecotone: Beacon Block Root

EIP-4788 introduces a "beacon block root" into the execution-layer block-header and EVM. This block root is an SSZ hash-tree-root of the consensus-layer contents of the previous consensus block.

With the adoption of EIP-4399 in the Bedrock upgrade the OP-Stack already includes the PREVRANDAO of L1. And thus with EIP-4788 the L1 beacon block root is made available.

For the Ecotone upgrade, this entails that:

• The parent_beacon_block_root of the L1 origin is now embedded in the L2 block header.
• The "Beacon roots contract" is deployed at Ecotone upgrade-time, or embedded at genesis if activated at genesis.
• The block state-transition process now includes the same special beacon-block-root EVM processing as L1 ethereum.

P2P Modifications

The Ethereum Node Record (ENR) for an Optimism execution node must contain an opel key-value pair where the key is opel and the value is a EIP-2124 fork id. The EL uses a different key from the CL in order to stop EL and CL nodes from connecting to each other.

Rollup Node Specification

Table of Contents

• Overview
• Driver
  • Derivation
• L2 Output RPC method
  • Structures
    • BlockID
    • L1BlockRef
    • L2BlockRef
    • SyncStatus
  • Output Method API
• Protocol Version tracking

Overview

The rollup node is the component responsible for deriving the L2 chain from L1 blocks (and their associated receipts).

The part of the rollup node that derives the L2 chain is called the rollup driver. This document is currently only concerned with the specification of the rollup driver.

Driver

The task of the driver in the rollup node is to manage the derivation process:

• Keep track of L1 head block
• Keep track of the L2 chain sync progress
• Iterate over the derivation steps as new inputs become available

Derivation

This process happens in three steps:

1. Select inputs from the L1 chain, on top of the last L2 block: a list of blocks, with transactions and associated data and receipts.
2. Read L1 information, deposits, and sequencing batches in order to generate payload attributes (essentially a block without output properties).
3. Pass the payload attributes to the execution engine, so that the L2 block (including output block properties) may be computed.

While this process is conceptually a pure function from the L1 chain to the L2 chain, it is in practice incremental. The L2 chain is extended whenever new L1 blocks are added to the L1 chain. Similarly, the L2 chain re-organizes whenever the L1 chain re-organizes.

For a complete specification of the L2 block derivation, refer to the L2 block derivation document.

L2 Output RPC method

The Rollup node has its own RPC method, optimism_outputAtBlock which returns a 32 byte hash corresponding to the L2 output root.

Structures

These define the types used by rollup node API methods. The types defined here are extended from the engine API specs.

BlockID

• hash: DATA, 32 Bytes
• number: QUANTITY, 64 Bits

L1BlockRef

• hash: DATA, 32 Bytes
• number: QUANTITY, 64 Bits
• parentHash: DATA, 32 Bytes
• timestamp: QUANTITY, 64 Bits

L2BlockRef

• hash: DATA, 32 Bytes
• number: QUANTITY, 64 Bits
• parentHash: DATA, 32 Bytes
• timestamp: QUANTITY, 64 Bits
• l1origin: BlockID
• sequenceNumber: QUANTITY, 64 Bits - distance to first block of epoch

SyncStatus

Represents a snapshot of the rollup driver.

• current_l1: Object - instance of L1BlockRef.
• current_l1_finalized: Object - instance of L1BlockRef.
• head_l1: Object - instance of L1BlockRef.
• safe_l1: Object - instance of L1BlockRef.
• finalized_l1: Object - instance of L1BlockRef.
• unsafe_l2: Object - instance of L2BlockRef.
• safe_l2: Object - instance of L2BlockRef.
• finalized_l2: Object - instance of L2BlockRef.
• pending_safe_l2: Object - instance of L2BlockRef.
• queued_unsafe_l2: Object - instance of L2BlockRef.

Output Method API

The input and return types here are as defined by the engine API specs.

• method: optimism_outputAtBlock
• params:
  1. blockNumber: QUANTITY, 64 bits - L2 integer block number.
• returns:
  1. version: DATA, 32 Bytes - the output root version number, beginning with 0.
  2. outputRoot: DATA, 32 Bytes - the output root.
  3. blockRef: Object - instance of L2BlockRef.
  4. withdrawalStorageRoot: 32 bytes - storage root of the L2toL1MessagePasser contract.
  5. stateRoot: DATA: 32 bytes - the state root.
  6. syncStatus: Object - instance of SyncStatus.

Protocol Version tracking

The rollup-node should monitor the recommended and required protocol version by monitoring the Protocol Version contract on L1, as specified in the Superchain Version Signaling specifications.

This can be implemented through polling in the Driver loop. After polling the Protocol Version, the rollup node SHOULD communicate it with the execution-engine through an engine_signalSuperchainV1 call.

The rollup node SHOULD warn the user when the recommended version is newer than the current version supported by the rollup node.

The rollup node SHOULD take safety precautions if it does not meet the required protocol version. This may include halting the engine, with consent of the rollup node operator.

Rollup-node P2P interface

Table of Contents

• Overview
• P2P configuration
  • Identification
  • Discv5
    • Consensus Layer Structure
  • LibP2P
    • Transport
    • Dialing
    • NAT
    • Peer management
    • Transport security
    • Protocol negotiation
    • Identify
    • Ping
    • Multiplexing
    • GossipSub
      • Content-based message identification
      • Message compression and limits
      • Message ID computation
    • Heartbeat and parameters
    • Topic configuration
    • Topic validation
• Gossip Topics
  • blocksv1
  • blocksv2
  • blocksv3
  • blocksv4
  • Block encoding
  • Block signatures
  • Block validation
    • Block processing
    • Block topic scoring parameters
• Req-Resp
  • payload_by_number

Overview

The rollup node has an optional peer-to-peer (P2P) network service to improve the latency between the view of sequencers and the rest of the network by bypassing the L1 in the happy case, without relying on a single centralized endpoint.

This also enables faster historical sync to be bootstrapped by providing block headers to sync towards, and only having to compare the L2 chain inputs to the L1 data as compared to processing everything one block at a time.

The rollup node will always prioritize L1 and reorganize to match the canonical chain. The L2 data retrieved via the P2P interface is strictly a speculative extension, also known as the "unsafe" chain, to improve the happy case performance.

This also means that P2P behavior is a soft-rule: nodes keep each other in check with scoring and eventual banning of malicious peers by identity or IP. Any behavior on the P2P layer does not affect the rollup security, at worst nodes rely on higher-latency data from L1 to serve.

In summary, the P2P stack looks like:

• Discovery to find peers: Discv5
• Connections, peering, transport security, multiplexing, gossip: LibP2P
• Application-layer publishing and validation of gossiped messages like L2 blocks.

This document only specifies the composition and configuration of these network libraries. These components have their own standards, implementations in Go/Rust/Java/Nim/JS/more, and are adopted by several other blockchains, most notably the L1 consensus layer (Eth2).

P2P configuration

Identification

Nodes have a separate network- and consensus-identity. The network identity is a secp256k1 key, used for both discovery and active LibP2P connections.

Common representations of network identity:

• PeerID: a LibP2P specific ID derived from the pubkey (through protobuf encoding, typing and hashing)
• NodeID: a Discv5 specific ID derived from the pubkey (through hashing, used in the DHT)
• Multi-address: an unsigned address, containing: IP, TCP port, PeerID
• ENR: a signed record used for discovery, containing: IP, TCP port, UDP port, signature (pubkey can be derived) and L2 network identification. Generally encoded in base64.

Discv5

Consensus Layer Structure

The Ethereum Node Record (ENR) for an Optimism rollup node must contain the following values, identified by unique keys:

• An IPv4 address (ip field) and/or IPv6 address (ip6 field).
• A TCP port (tcp field) representing the local libp2p listening port.
• A UDP port (udp field) representing the local discv5 listening port.
• An OpStack (opstack field) L2 network identifier

The opstack value is encoded as a single RLP bytes value, the concatenation of:

• chain ID (unsigned varint)
• fork ID (unsigned varint)

Note that DiscV5 is a shared DHT (Distributed Hash Table): the L1 consensus and execution nodes, as well as testnet nodes, and even external IOT nodes, all communicate records in this large common DHT. This makes it more difficult to censor the discovery of node records.

The discovery process in Optimism is a pipeline of node records:

1. Fill the table with FINDNODES if necessary (Performed by Discv5 library)
2. Pull additional records with searches to random Node IDs if necessary (e.g. iterate RandomNodes() in Go implementation)
3. Pull records from the DiscV5 module when looking for peers
4. Check if the record contains the opstack entry, verify it matches the chain ID and current or future fork number
5. If not already connected, and not recently disconnected or put on deny-list, attempt to dial.

LibP2P

Transport

TCP transport. Additional transports are supported by LibP2P, but not required.

Dialing

Nodes should be publicly dialable, not rely on relay extensions, and able to dial both IPv4 and IPv6.

NAT

The listening endpoint must be publicly facing, but may be configured behind a NAT. LibP2P will use PMP / UPNP based techniques to track the external IP of the node. It is recommended to disable the above if the external IP is static and configured manually.

Peer management

The default is to maintain a peer count with a tide-system based on active peer count:

• At "low tide" the node starts to actively search for additional peer connections.
• At "high tide" the node starts to prune active connections, except those that are marked as trusted or have a grace period.

Peers will have a grace period for a configurable amount of time after joining. In an emergency, when memory runs low, the node should start pruning more aggressively.

Peer records can be persisted to disk to quickly reconnect with known peers after restarting the rollup node.

The discovery process feeds the peerstore with peer records to connect to, tagged with a time-to-live (TTL). The current P2P processes do not require selective topic-specific peer connections, other than filtering for the basic network participation requirement.

Peers may be banned if their performance score is too low, or if an objectively malicious action was detected.

Banned peers will be persisted to the same data-store as the peerstore records.

TODO: the connection gater does currently not gate by IP address on the dial Accept-callback.

Transport security

Libp2p-noise, XX handshake, with the secp256k1 P2P identity, as popularized in Eth2. The TLS option is available as well, but noise should be prioritized in negotiation.

Protocol negotiation

Multistream-select 1.0 (/multistream/1.0.0) is an interactive protocol used to negotiate sub-protocols supported in LibP2P peers. Multistream-select 2.0 may be used in the future.

Identify

LibP2P offers a minimal identification module to share client version and programming language. This is optional and can be disabled for enhanced privacy. It also includes the same protocol negotiation information, which can speed up initial connections.

Ping

LibP2P includes a simple ping protocol to track latency between connections. This should be enabled to help provide insight into the network health.

Multiplexing

For async communication over different channels over the same connection, multiplexing is used. mplex (/mplex/6.7.0) is required, and yamux (/yamux/1.0.0) is recommended but optional

GossipSub

GossipSub 1.1 (/meshsub/1.1.0, i.e. with peer-scoring extension) is a pubsub protocol for mesh-networks, deployed on L1 consensus (Eth2) and other protocols such as Filecoin, offering lots of customization options.

Content-based message identification

Messages are deduplicated, and filtered through application-layer signature verification. Thus origin-stamping is disabled and published messages must only contain application data, enforced through a StrictNoSign Signature Policy

This provides greater privacy, and allows sequencers (consensus identity) to maintain multiple network identities for redundancy.

Message compression and limits

The application contents are compressed with snappy single-block-compression (as opposed to frame-compression), and constrained to 10 MiB.

Message ID computation

Same as L1, with recognition of compression:

• If message.data has a valid snappy decompression, set message-id to the first 20 bytes of the SHA256 hash of the concatenation of MESSAGE_DOMAIN_VALID_SNAPPY with the snappy decompressed message data, i.e. SHA256(MESSAGE_DOMAIN_VALID_SNAPPY + snappy_decompress(message.data))[:20].
• Otherwise, set message-id to the first 20 bytes of the SHA256 hash of the concatenation of MESSAGE_DOMAIN_INVALID_SNAPPY with the raw message data, i.e. SHA256(MESSAGE_DOMAIN_INVALID_SNAPPY + message.data)[:20].

Heartbeat and parameters

GossipSub parameters:

• D (topic stable mesh target count): 8
• D_low (topic stable mesh low watermark): 6
• D_high (topic stable mesh high watermark): 12
• D_lazy (gossip target): 6
• heartbeat_interval (interval of heartbeat, in seconds): 0.5
• fanout_ttl (ttl for fanout maps for topics we are not subscribed to but have published to, in seconds): 24
• mcache_len (number of windows to retain full messages in cache for IWANT responses): 12
• mcache_gossip (number of windows to gossip about): 3
• seen_ttl (number of heartbeat intervals to retain message IDs): 130 (= 65 seconds)

Notable differences from L1 consensus (Eth2):

• seen_ttl does not need to cover a full L1 epoch (6.4 minutes), but rather just a small window covering latest blocks
• fanout_ttl: adjusted to lower than seen_ttl
• mcache_len: a larger number of heartbeats can be retained since the gossip is much less noisy.
• heartbeat_interval: faster interval to reduce latency, bandwidth should still be reasonable since there are far fewer messages to gossip about each interval than on L1 which uses an interval of 0.7 seconds.

Topic configuration

Topics have string identifiers and are communicated with messages and subscriptions. /optimism/chain_id/hardfork_version/Name

• chain_id: replace with decimal representation of chain ID
• hardfork_version: replace with decimal representation of hardfork, starting at 0
• Name: topic application-name

Note that the topic encoding depends on the topic, unlike L1, since there are less topics, and all are snappy-compressed.

Topic validation

To ensure only valid messages are relayed, and malicious peers get scored based on application behavior, an extended validator checks the message before it is relayed or processed. The extended validator emits one of the following validation signals:

• ACCEPT valid, relayed to other peers and passed to local topic subscriber
• IGNORE scored like inactivity, message is dropped and not processed
• REJECT score penalties, message is dropped

Gossip Topics

Listed below are the topics for distributing blocks to other nodes faster than proxying through L1 would. These are:

blocksv1

Pre-Canyon/Shanghai blocks are broadcast on /optimism/<chainId>/0/blocks.

blocksv2

Canyon/Delta blocks are broadcast on /optimism/<chainId>/1/blocks.

blocksv3

Ecotone blocks are broadcast on /optimism/<chainId>/2/blocks.

blocksv4

Isthmus blocks are broadcast on /optimism/<chainId>/3/blocks.

Block encoding

A block is structured as the concatenation of:

• V1 and V2 topics
  • signature: A secp256k1 signature, always 65 bytes, r (uint256), s (uint256), y_parity (uint8)
  • payload: A SSZ-encoded ExecutionPayload, always the remaining bytes.
• V3 topic
  • signature: A secp256k1 signature, always 65 bytes, r (uint256), s (uint256), y_parity (uint8)
  • parentBeaconBlockRoot: L1 origin parent beacon block root, always 32 bytes
  • payload: A SSZ-encoded ExecutionPayload, always the remaining bytes.
• V4 topic
  • signature: A secp256k1 signature, always 65 bytes, r (uint256), s (uint256), y_parity (uint8)
  • parentBeaconBlockRoot: L1 origin parent beacon block root, always 32 bytes
  • withdrawalsRoot: L2 withdrawals root, always 32 bytes.
  • payload: A SSZ-encoded ExecutionPayload, always the remaining bytes.

All topics use Snappy block-compression (i.e. no snappy frames): the above needs to be compressed after encoding, and decompressed before decoding.

Block signatures

The signature is a secp256k1 signature, and signs over a message: keccak256(domain ++ chain_id ++ payload_hash), where:

• domain is 32 bytes, reserved for message types and versioning info. All zero for this signature.
• chain_id is a big-endian encoded uint256.
• payload_hash is keccak256(payload), where payload is:
  • the payload in V1 and V2,
  • parentBeaconBlockRoot ++ payload in V3.
  • parentBeaconBlockRoot ++ withdrawalsRoot ++ payload in V4.

The secp256k1 signature must have y_parity = 1 or 0, the chain_id is already signed over.

Block validation

An extended-validator checks the incoming messages as follows, in order of operation:

• [REJECT] if the compression is not valid
• [REJECT] if the block encoding is not valid
• [REJECT] if the payload.timestamp is older than 60 seconds in the past (graceful boundary for worst-case propagation and clock skew)
• [REJECT] if the payload.timestamp is more than 5 seconds into the future
• [REJECT] if the block_hash in the payload is not valid
• [REJECT] if the block is on the V1 topic and has withdrawals
• [REJECT] if the block is on the V1 topic and has a withdrawals list
• [REJECT] if the block is on a topic >= V2 and does not have an empty withdrawals list
• [REJECT] if the block is on a topic <= V2 and has a blob gas-used value set
• [REJECT] if the block is on a topic <= V2 and has an excess blob gas value set
• [REJECT] if the block is on a topic >= V3 and has a blob gas-used value that is not zero
• [REJECT] if the block is on a topic >= V3 and has an excess blob gas value that is not zero
• [REJECT] if the block is on a topic <= V2 and the parent beacon block root is not nil
• [REJECT] if the block is on a topic >= V3 and the parent beacon block root is nil
• [REJECT] if the block is on a topic >= V3 and the l2 withdrawals root is nil
• [REJECT] if more than 5 different blocks have been seen with the same block height
• [IGNORE] if the block has already been seen
• [REJECT] if the signature by the sequencer is not valid
• Mark the block as seen for the given block height

The block is signed by the corresponding sequencer, to filter malicious messages. The sequencer model is singular but may change to multiple sequencers in the future. A default sequencer pubkey is distributed with rollup nodes and should be configurable.

Note that blocks that a block may still be propagated even if the L1 already confirmed a different block. The local L1 view of the node may be wrong, and the time and signature validation will prevent spam. Hence, calling into the execution engine with a block lookup every propagation step is not worth the added delay.

Block processing

A node may apply the block to their local engine ahead of L1 availability, if it ensures that:

• The application of the block is reversible, in case of a conflict with delayed L1 information
• The subsequent forkchoice-update ensures this block is recognized as "unsafe" (see fork choice updated)

Block topic scoring parameters

TODO: GossipSub per-topic scoring to fine-tune incentives for ideal propagation delay and bandwidth usage.

Req-Resp

The op-node implements a similar request-response encoding for its sync protocols as the L1 ethereum Beacon-Chain. See L1 P2P-interface req-resp specification and Altair P2P update.

However, the protocol is simplified, to avoid several issues seen in L1:

• Error strings in responses, if there is any alternative response, should not need to be compressed or have an artificial global length limit.
• Payload lengths should be fixed-length: byte-by-byte uvarint reading from the underlying stream is undesired.
• <context-bytes> are relaxed to encode a uint32, rather than a beacon-chain ForkDigest.
• Payload-encoding may change per hardfork, so is not part of the protocol-ID.
• Usage of response-chunks is specific to the req-resp method: most basic req-resp does not need chunked responses.
• Compression is encouraged to be part of the payload-encoding, specific to the req-resp method, where necessary: pings and such do not need streaming frame compression etc.

And the protocol ID format follows the same scheme as L1, except the trailing encoding schema part, which is now message-specific:

/ProtocolPrefix/MessageName/SchemaVersion/

The req-resp protocols served by the op-node all have /ProtocolPrefix set to /opstack/req.

Individual methods may include the chain ID as part of the /MessageName segment, so it's immediately clear which chain the method applies to, if the communication is chain-specific. Other methods may include chain-information in the request and/or response data, such as the ForkDigest <context-bytes> in L1 beacon chain req-resp protocols.

Each segment starts with a /, and may contain multiple /, and the final protocol ID is suffixed with a /.

payload_by_number

This is an optional chain syncing method, to request/serve execution payloads by number. This serves as a method to fill gaps upon missed gossip, and sync short to medium ranges of unsafe L2 blocks.

Protocol ID: /opstack/req/payload_by_number/<chain-id>/0/

• /MessageName is /block_by_number/<chain-id> where <chain-id> is set to the op-node L2 chain ID.
• /SchemaVersion is /0

Request format: <num>: a little-endian uint64 - the block number to request.

Response format: <response> = <res><version><payload>

• <res> is a byte code describing the result.
  • 0 on success, <version><payload> should follow.
  • 1 if valid request, but unavailable payload.
  • 2 if invalid request
  • 3+ if other error
  • The >= 128 range is reserved for future use.
• <version> is a little-endian uint32, identifying the response type (fork-specific)
• <payload> is an encoded block, read till stream EOF.

The input of <response> should be limited, as well as any generated decompressed output, to avoid unexpected resource usage or zip-bomb type attacks. A 10 MB limit is recommended, to ensure all blocks may be synced. Implementations may opt for a different limit, since this sync method is optional.

<version> list:

• 0: SSZ-encoded ExecutionPayload, with Snappy framing compression, matching the ExecutionPayload SSZ definition of the L1 Merge, L2 Bedrock and L2 Regolith, L2 Canyon versions.
• 1: SSZ-encoded ExecutionPayloadEnvelope with Snappy framing compression, matching the ExecutionPayloadEnvelope SSZ definition of the L2 Ecotone version.
• 2: SSZ-encoded ExecutionPayload with Snappy framing compression, matching the ExecutionPayload SSZ definition of the L2 Isthmus version.

The request is by block-number, enabling parallel fetching of a chain across many peers.

A res = 0 response should be verified to:

• Have a block-number matching the requested block number.
• Have a consistent blockhash w.r.t. the other block contents.
• Build towards a known canonical block.
  • This can be verified by checking if the parent-hash of a previous trusted canonical block matches that of the verified hash of the retrieved block.
  • For unsafe blocks this may be relaxed to verification against the parent-hash of any previously trusted block:
    • The gossip validation process limits the amount of blocks that may be trusted to sync towards.
    • The unsafe blocks should be queued for processing, the latest received L2 unsafe blocks should always override any previous chain, until the final L2 chain can be reproduced from L1 data.

A res > 0 response code should not be accepted. The result code is helpful for debugging, but the client should regard any error like any other unanswered request, as the responding peer cannot be trusted.

L2 Chain Derivation Specification

Table of Contents

• Overview
  • Eager Block Derivation
  • Protocol Parameters
• Batch Submission
  • Sequencing & Batch Submission Overview
  • Batch Submission Wire Format
    • Batcher Transaction Format
    • Frame Format
    • Channel Format
    • Batch Format
• Architecture
  • L2 Chain Derivation Pipeline
    • L1 Traversal
    • L1 Retrieval
    • Frame Queue
    • Channel Bank
      • Pruning
      • Timeouts
      • Reading
      • Loading frames
    • Channel Reader (Batch Decoding)
    • Batch Queue
    • Payload Attributes Derivation
    • Engine Queue
      • Engine API usage
        • Bedrock, Canyon, Delta: API Usage
        • Ecotone: API Usage
      • Forkchoice synchronization
      • L1-consolidation: payload attributes matching
      • L1-sync: payload attributes processing
      • Processing unsafe payload attributes
    • Resetting the Pipeline
      • Finding the sync starting point
      • Resetting derivation stages
      • About reorgs Post-Merge
• Deriving Payload Attributes
  • Deriving the Transaction List
    • Network upgrade automation transactions
      • Ecotone
        • L1Block Deployment
        • GasPriceOracle Deployment
        • L1Block Proxy Update
        • GasPriceOracle Proxy Update
        • GasPriceOracle Enable Ecotone
        • Beacon block roots contract deployment (EIP-4788)
  • Building Individual Payload Attributes

Overview

Note the following assumes a single sequencer and batcher. In the future, the design will be adapted to accommodate multiple such entities.

L2 chain derivation — deriving L2 blocks from L1 data — is one of the main responsibilities of the rollup node, both in validator mode, and in sequencer mode (where derivation acts as a sanity check on sequencing, and enables detecting L1 chain re-organizations).

The L2 chain is derived from the L1 chain. In particular, each L1 block following L2 chain inception is mapped to a sequencing epoch comprising at least one L2 block. Each L2 block belongs to exactly one epoch, and we call the corresponding L1 block its L1 origin. The epoch's number equals that of its L1 origin block.

To derive the L2 blocks of epoch number E, we need the following inputs:

• L1 blocks in the range [E, E + SWS), called the sequencing window of the epoch, and SWS the sequencing window size. (Note that sequencing windows overlap.)
• Batcher transactions from blocks in the sequencing window.
  • These transactions allow us to reconstruct the epoch's sequencer batches, each of which will produce one L2 block. Note that:
    • The L1 origin will never contain any data needed to construct sequencer batches since each batch must contain the L1 origin hash.
    • An epoch may have no sequencer batches.
• Deposits made in the L1 origin (in the form of events emitted by the deposit contract).
• L1 block attributes from the L1 origin (to derive the L1 attributes deposited transaction).
• The state of the L2 chain after the last L2 block of the previous epoch, or the L2 genesis state if E is the first epoch.

To derive the whole L2 chain from scratch, we start with the L2 genesis state and the L2 genesis block as the first L2 block. We then derive L2 blocks from each epoch in order, starting at the first L1 block following L2 chain inception. Refer to the Architecture section for more information on how we implement this in practice. The L2 chain may contain pre-Bedrock history, but the L2 genesis here refers to the Bedrock L2 genesis block.

Each L2 block with origin l1_origin is subject to the following constraints (whose values are denominated in seconds):

• block.timestamp = prev_l2_timestamp + l2_block_time

  • prev_l2_timestamp is the timestamp of the L2 block immediately preceding this one. If there is no preceding block, then this is the genesis block, and its timestamp is explicitly specified.
  • l2_block_time is a configurable parameter of the time between L2 blocks (2s on Optimism).

• l1_origin.timestamp <= block.timestamp <= max_l2_timestamp, where

  • max_l2_timestamp = max(l1_origin.timestamp + max_sequencer_drift, prev_l2_timestamp + l2_block_time)
    • max_sequencer_drift is a configurable parameter that bounds how far the sequencer can get ahead of the L1.

Finally, each epoch must have at least one L2 block.

The first constraint means there must be an L2 block every l2_block_time seconds following L2 chain inception.

The second constraint ensures that an L2 block timestamp never precedes its L1 origin timestamp, and is never more than max_sequencer_drift ahead of it, except only in the unusual case where it might prohibit an L2 block from being produced every l2_block_time seconds. (Such cases might arise for example under a proof-of-work L1 that sees a period of rapid L1 block production.) In either case, the sequencer enforces len(batch.transactions) == 0 while max_sequencer_drift is exceeded. See Batch Queue for more details.

The final requirement that each epoch must have at least one L2 block ensures that all relevant information from the L1 (e.g. deposits) is represented in the L2, even if it has no sequencer batches.

Post-merge, Ethereum has a fixed 12s block time, though some slots can be skipped. Under a 2s L2 block time, we thus expect each epoch to typically contain 12/2 = 6 L2 blocks. The sequencer will however produce bigger epochs in order to maintain liveness in case of either a skipped slot on the L1 or a temporary loss of connection to it. For the lost connection case, smaller epochs might be produced after the connection was restored to keep L2 timestamps from drifting further and further ahead.

Eager Block Derivation

Deriving an L2 block requires that we have constructed its sequencer batch and derived all L2 blocks and state updates prior to it. This means we can typically derive the L2 blocks of an epoch eagerly without waiting on the full sequencing window. The full sequencing window is required before derivation only in the very worst case where some portion of the sequencer batch for the first block of the epoch appears in the very last L1 block of the window. Note that this only applies to block derivation. Sequencer batches can still be derived and tentatively queued without deriving blocks from them.

Protocol Parameters

The following table gives an overview of some protocol parameters, and how they are affected by protocol upgrades.

Batch Submission

Sequencing & Batch Submission Overview

The sequencer accepts L2 transactions from users. It is responsible for building blocks out of these. For each such block, it also creates a corresponding sequencer batch. It is also responsible for submitting each batch to a data availability provider (e.g. Ethereum calldata), which it does via its batcher component.

The difference between an L2 block and a batch is subtle but important: the block includes an L2 state root, whereas the batch only commits to transactions at a given L2 timestamp (equivalently: L2 block number). A block also includes a reference to the previous block (*).

(*) This matters in some edge case where a L1 reorg would occur and a batch would be reposted to the L1 chain but not the preceding batch, whereas the predecessor of an L2 block cannot possibly change.

This means that even if the sequencer applies a state transition incorrectly, the transactions in the batch will still be considered part of the canonical L2 chain. Batches are still subject to validity checks (i.e. they have to be encoded correctly), and so are individual transactions within the batch (e.g. signatures have to be valid). Invalid batches and invalid individual transactions within an otherwise valid batch are discarded by correct nodes.

If the sequencer applies a state transition incorrectly and posts an output root, then this output root will be incorrect. The incorrect output root will be challenged by a fault proof, then replaced by a correct output root for the existing sequencer batches.

Refer to the Batch Submission specification for more information.

Batch Submission Wire Format

Batch submission is closely tied to L2 chain derivation because the derivation process must decode the batches that have been encoded for the purpose of batch submission.

The batcher submits batcher transactions to a data availability provider. These transactions contain one or multiple channel frames, which are chunks of data belonging to a channel.

A channel is a sequence of sequencer batches (for any L2 blocks) compressed together. The reason to group multiple batches together is simply to obtain a better compression rate, hence reducing data availability costs.

Channels might be too large to fit in a single batcher transaction, hence we need to split it into chunks known as channel frames. A single batcher transaction can also carry multiple frames (belonging to the same or to different channels).

This design gives use the maximum flexibility in how we aggregate batches into channels, and split channels over batcher transactions. It notably allows us to maximize data utilization in a batcher transaction: for instance it allows us to pack the final (small) frame of one channel with one or more frames from the next channel.

Also note that we use a streaming compression scheme, and we do not need to know how many batches a channel will end up containing when we start a channel, or even as we send the first frames in the channel.

And by splitting channels across multiple data transactions, the L2 can have larger block data than the data-availability layer may support.

All of this is illustrated in the following diagram. Explanations below.

The first line represents L1 blocks with their numbers. The boxes under the L1 blocks represent batcher transactions included within the block. The squiggles under the L1 blocks represent deposits (more specifically, events emitted by the deposit contract).

Each colored chunk within the boxes represents a channel frame. So A and B are channels whereas A0, A1, B0, B1, B2 are frames. Notice that:

• multiple channels are interleaved
• frames do not need to be transmitted in order
• a single batcher transaction can carry frames from multiple channels

In the next line, the rounded boxes represent individual sequencer batches that were extracted from the channels. The four blue/purple/pink were derived from channel A while the other were derived from channel B. These batches are here represented in the order they were decoded from batches (in this case B is decoded first).

Note The caption here says "Channel B was seen first and will be decoded into batches first", but this is not a requirement. For instance, it would be equally acceptable for an implementation to peek into the channels and decode the one that contains the oldest batches first.

The rest of the diagram is conceptually distinct from the first part and illustrates L2 chain derivation after the channels have been reordered.

The first line shows batcher transactions. Note that in this case, there exists an ordering of the batches that makes all frames within the channels appear contiguously. This is not true in general. For instance, in the second transaction, the position of A1 and B0 could have been inverted for exactly the same result — no changes needed in the rest of the diagram.

The second line shows the reconstructed channels in proper order. The third line shows the batches extracted from the channel. Because the channels are ordered and the batches within a channel are sequential, this means the batches are ordered too. The fourth line shows the L2 block derived from each batch. Note that we have a 1-1 batch to block mapping here but, as we'll see later, empty blocks that do not map to batches can be inserted in cases where there are "gaps" in the batches posted on L1.

The fifth line shows the L1 attributes deposited transaction which, within each L2 block, records information about the L1 block that matches the L2 block's epoch. The first number denotes the epoch/L1x number, while the second number (the "sequence number") denotes the position within the epoch.

Finally, the sixth line shows user-deposited transactions derived from the deposit contract event mentioned earlier.

Note the 101-0 L1 attributes transaction on the bottom right of the diagram. Its presence there is only possible if frame B2 indicates that it is the last frame within the channel and (2) no empty blocks must be inserted.

The diagram does not specify the sequencing window size in use, but from this we can infer that it must be at least 4 blocks, because the last frame of channel A appears in block 102, but belong to epoch 99.

As for the comment on "security types", it explains the classification of blocks as used on L1 and L2.

• Unsafe L2 blocks:
• Safe L2 blocks:
• Finalized L2 blocks: refer to block that have been derived from finalized L1 data.

These security levels map to the headBlockHash, safeBlockHash and finalizedBlockHash values transmitted when interacting with the execution-engine API.

Batcher Transaction Format

Batcher transactions are encoded as version_byte ++ rollup_payload (where ++ denotes concatenation).

Unknown versions make the batcher transaction invalid (it must be ignored by the rollup node). All frames in a batcher transaction must be parseable. If any one frame fails to parse, the all frames in the transaction are rejected.

Batch transactions are authenticated by verifying that the to address of the transaction matches the batch inbox address, and the from address matches the batch-sender address in the system configuration at the time of the L1 block that the transaction data is read from.

Frame Format

A channel frame is encoded as:

frame = channel_id ++ frame_number ++ frame_data_length ++ frame_data ++ is_last

channel_id        = bytes16
frame_number      = uint16
frame_data_length = uint32
frame_data        = bytes
is_last           = bool

Where uint32 and uint16 are all big-endian unsigned integers. Type names should be interpreted to and encoded according to the Solidity ABI.

All data in a frame is fixed-size, except the frame_data. The fixed overhead is 16 + 2 + 4 + 1 = 23 bytes. Fixed-size frame metadata avoids a circular dependency with the target total data length, to simplify packing of frames with varying content length.

where:

• channel_id is an opaque identifier for the channel. It should not be reused and is suggested to be random; however, outside of timeout rules, it is not checked for validity
• frame_number identifies the index of the frame within the channel
• frame_data_length is the length of frame_data in bytes. It is capped to 1,000,000 bytes.
• frame_data is a sequence of bytes belonging to the channel, logically after the bytes from the previous frames
• is_last is a single byte with a value of 1 if the frame is the last in the channel, 0 if there are frames in the channel. Any other value makes the frame invalid (it must be ignored by the rollup node).

Channel Format

A channel is encoded by applying a streaming compression algorithm to a list of batches:

encoded_batches = []
for batch in batches:
    encoded_batches ++ batch.encode()
rlp_batches = rlp_encode(encoded_batches)

where:

• batches is the input, a sequence of batches each with a byte-encoder function .encode() as per the next section ("Batch Encoding")
• encoded_batches is a byte array: the concatenation of the encoded batches
• rlp_batches is the rlp encoding of the concatenated encoded batches

channel_encoding = zlib_compress(rlp_batches)

where zlib_compress is the ZLIB algorithm (as specified in RFC-1950) with no dictionary.

The Fjord upgrade introduces an additional versioned channel encoding format to support alternate compression algorithms.

When decompressing a channel, we limit the amount of decompressed data to MAX_RLP_BYTES_PER_CHANNEL (defined in the Protocol Parameters table), in order to avoid "zip-bomb" types of attack (where a small compressed input decompresses to a humongous amount of data). If the decompressed data exceeds the limit, things proceeds as though the channel contained only the first MAX_RLP_BYTES_PER_CHANNEL decompressed bytes. The limit is set on RLP decoding, so all batches that can be decoded in MAX_RLP_BYTES_PER_CHANNEL will be accepted even if the size of the channel is greater than MAX_RLP_BYTES_PER_CHANNEL. The exact requirement is that length(input) <= MAX_RLP_BYTES_PER_CHANNEL.

While the above pseudocode implies that all batches are known in advance, it is possible to perform streaming compression and decompression of RLP-encoded batches. This means it is possible to start including channel frames in a batcher transaction before we know how many batches (and how many frames) the channel will contain.

Batch Format

Recall that a batch contains a list of transactions to be included in a specific L2 block.

A batch is encoded as batch_version ++ content, where content depends on the batch_version. Prior to the Delta upgrade, batches all have batch_version 0 and are encoded as described below.

where:

• batch_version is a single byte, prefixed before the RLP contents, alike to transaction typing.
• rlp_encode is a function that encodes a batch according to the RLP format, and [x, y, z] denotes a list containing items x, y and z
• parent_hash is the block hash of the previous L2 block
• epoch_number and epoch_hash are the number and hash of the L1 block corresponding to the sequencing epoch of the L2 block
• timestamp is the timestamp of the L2 block
• transaction_list is an RLP-encoded list of EIP-2718 encoded transactions.

The Delta upgrade introduced an additional batch type, span batches.

Unknown versions make the batch invalid (it must be ignored by the rollup node), as do malformed contents.

Note if the batch version and contents can be RLP decoded correctly but extra content exists beyond the batch, the additional data may be ignored during parsing. Data between RLP encoded batches may not be ignored (as they are seen as malformed batches), but if a batch can be fully described by the RLP decoding, extra content does not invalidate the decoded batch.

The epoch_number and the timestamp must also respect the constraints listed in the Batch Queue section, otherwise the batch is considered invalid and will be ignored.

Architecture

The above primarily describes the general encodings used in L2 chain derivation, primarily how batches are encoded within batcher transactions.

This section describes how the L2 chain is produced from the L1 batches using a pipeline architecture.

A verifier may implement this differently, but must be semantically equivalent to not diverge from the L2 chain.

L2 Chain Derivation Pipeline

Our architecture decomposes the derivation process into a pipeline made up of the following stages:

1. L1 Traversal
2. L1 Retrieval
3. Frame Queue
4. Channel Bank
5. Channel Reader (Batch Decoding)
6. Batch Queue
7. Payload Attributes Derivation
8. Engine Queue

The data flows from the start (outer) of the pipeline towards the end (inner). From the innermost stage the data is pulled from the outermost stage.

However, data is processed in reverse order. Meaning that if there is any data to be processed in the last stage, it will be processed first. Processing proceeds in "steps" that can be taken at each stage. We try to take as many steps as possible in the last (most inner) stage before taking any steps in its outer stage, etc.

This ensures that we use the data we already have before pulling more data and minimizes the latency of data traversing the derivation pipeline.

Each stage can maintain its own inner state as necessary. In particular, each stage maintains a L1 block reference (number + hash) to the latest L1 block such that all data originating from previous blocks has been fully processed, and the data from that block is being or has been processed. This allows the innermost stage to account for finalization of the L1 data-availability used to produce the L2 chain, to reflect in the L2 chain forkchoice when the L2 chain inputs become irreversible.

Let's briefly describe each stage of the pipeline.

L1 Traversal

In the L1 Traversal stage, we simply read the header of the next L1 block. In normal operations, these will be new L1 blocks as they get created, though we can also read old blocks while syncing, or in case of an L1 re-org.

Upon traversal of the L1 block, the system configuration copy used by the L1 retrieval stage is updated, such that the batch-sender authentication is always accurate to the exact L1 block that is read by the stage.

L1 Retrieval

In the L1 Retrieval stage, we read the block we get from the outer stage (L1 traversal), and extract data from its batcher transactions. A batcher transaction is one with the following properties:

• The to field is equal to the configured batcher inbox address.

• The sender, as recovered from the transaction signature (v, r, and s), is the batcher address loaded from the system config matching the L1 block of the data.

Each batcher transaction is versioned and contains a series of channel frames to be read by the Frame Queue, see Batch Submission Wire Format. Each batcher transaction in the block is processed in the order they appear in the block by passing its calldata on to the next phase.

Frame Queue

The Frame Queue buffers one data-transaction at a time, decoded into channel frames, to be consumed by the next stage. See Batcher transaction format and Frame format specifications.

Channel Bank

The Channel Bank stage is responsible for managing buffering from the channel bank that was written to by the L1 retrieval stage. A step in the channel bank stage tries to read data from channels that are "ready".

Channels are currently fully buffered until read or dropped, streaming channels may be supported in a future version of the ChannelBank.

To bound resource usage, the Channel Bank prunes based on channel size, and times out old channels.

Channels are recorded in FIFO order in a structure called the channel queue. A channel is added to the channel queue the first time a frame belonging to the channel is seen.

Pruning

After successfully inserting a new frame, the ChannelBank is pruned: channels are dropped in FIFO order, until total_size <= MAX_CHANNEL_BANK_SIZE, where:

• total_size is the sum of the sizes of each channel, which is the sum of all buffered frame data of the channel, with an additional frame-overhead of 200 bytes per frame.
• MAX_CHANNEL_BANK_SIZE is a protocol constant defined in the Protocol Parameters table.

Timeouts

The L1 origin that the channel was opened in is tracked with the channel as channel.open_l1_block, and determines the maximum span of L1 blocks that the channel data is retained for, before being pruned.

A channel is timed out if: current_l1_block.number > channel.open_l1_block.number + CHANNEL_TIMEOUT, where:

• current_l1_block is the L1 origin that the stage is currently traversing.
• CHANNEL_TIMEOUT is a rollup-configurable, expressed in number of L1 blocks.

New frames for timed-out channels are dropped instead of buffered.

Reading

Upon reading, while the first opened channel is timed-out, remove it from the channel-bank.

Prior to the Canyon network upgrade, once the first opened channel, if any, is not timed-out and is ready, then it is read and removed from the channel-bank. After the Canyon network upgrade, the entire channel bank is scanned in FIFO order (by open time) & the first ready (i.e. not timed-out) channel will be returned.

The canyon behavior will activate when frames from a L1 block whose timestamp is greater than or equal to the canyon time first enter the channel queue.

A channel is ready if:

• The channel is closed
• The channel has a contiguous sequence of frames until the closing frame

If no channel is ready, the next frame is read and ingested into the channel bank.

Loading frames

When a channel ID referenced by a frame is not already present in the Channel Bank, a new channel is opened, tagged with the current L1 block, and appended to the channel-queue.

Frame insertion conditions:

• New frames matching timed-out channels that have not yet been pruned from the channel-bank are dropped.
• Duplicate frames (by frame number) for frames that have not been pruned from the channel-bank are dropped.
• Duplicate closes (new frame is_last == 1, but the channel has already seen a closing frame and has not yet been pruned from the channel-bank) are dropped.

If a frame is closing (is_last == 1) any existing higher-numbered frames are removed from the channel.

Note that while this allows channel IDs to be reused once they have been pruned from the channel-bank, it is recommended that batcher implementations use unique channel IDs.

Channel Reader (Batch Decoding)

In this stage, we decompress the channel we pull from the last stage, and then parse batches from the decompressed byte stream.

See Channel Format and Batch Format for decompression and decoding specification.

Batch Queue

During the Batch Buffering stage, we reorder batches by their timestamps. If batches are missing for some time slots and a valid batch with a higher timestamp exists, this stage also generates empty batches to fill the gaps.

Batches are pushed to the next stage whenever there is one sequential batch directly following the timestamp of the current safe L2 head (the last block that can be derived from the canonical L1 chain). The parent hash of the batch must also match the hash of the current safe L2 head.

Note that the presence of any gaps in the batches derived from L1 means that this stage will need to buffer for a whole sequencing window before it can generate empty batches (because the missing batch(es) could have data in the last L1 block of the window in the worst case).

A batch can have 4 different forms of validity:

• drop: the batch is invalid, and will always be in the future, unless we reorg. It can be removed from the buffer.
• accept: the batch is valid and should be processed.
• undecided: we are lacking L1 information until we can proceed batch filtering.
• future: the batch may be valid, but cannot be processed yet and should be checked again later.

The batches are processed in order of the inclusion on L1: if multiple batches can be accept-ed the first is applied. An implementation can defer future batches a later derivation step to reduce validation work.

The batches validity is derived as follows:

Definitions:

• batch as defined in the Batch format section.
• epoch = safe_l2_head.l1_origin a L1 origin coupled to the batch, with properties: number (L1 block number), hash (L1 block hash), and timestamp (L1 block timestamp).
• inclusion_block_number is the L1 block number when batch was first fully derived, i.e. decoded and output by the previous stage.
• next_timestamp = safe_l2_head.timestamp + block_time is the expected L2 timestamp the next batch should have, see block time information.
• next_epoch may not be known yet, but would be the L1 block after epoch if available.
• batch_origin is either epoch or next_epoch, depending on validation.

Note that processing of a batch can be deferred until batch.timestamp <= next_timestamp, since future batches will have to be retained anyway.

Rules, in validation order:

• batch.timestamp > next_timestamp -> future: i.e. the batch must be ready to process.
• batch.timestamp < next_timestamp -> drop: i.e. the batch must not be too old.
• batch.parent_hash != safe_l2_head.hash -> drop: i.e. the parent hash must be equal to the L2 safe head block hash.
• batch.epoch_num + sequence_window_size < inclusion_block_number -> drop: i.e. the batch must be included timely.
• batch.epoch_num < epoch.number -> drop: i.e. the batch origin is not older than that of the L2 safe head.
• batch.epoch_num == epoch.number: define batch_origin as epoch.
• batch.epoch_num == epoch.number+1:
  • If next_epoch is not known -> undecided: i.e. a batch that changes the L1 origin cannot be processed until we have the L1 origin data.
  • If known, then define batch_origin as next_epoch
• batch.epoch_num > epoch.number+1 -> drop: i.e. the L1 origin cannot change by more than one L1 block per L2 block.
• batch.epoch_hash != batch_origin.hash -> drop: i.e. a batch must reference a canonical L1 origin, to prevent batches from being replayed onto unexpected L1 chains.
• batch.timestamp < batch_origin.time -> drop: enforce the min L2 timestamp rule.
• batch.timestamp > batch_origin.time + max_sequencer_drift: enforce the L2 timestamp drift rule, but with exceptions to preserve above min L2 timestamp invariant:
  • len(batch.transactions) == 0:
    • epoch.number == batch.epoch_num: this implies the batch does not already advance the L1 origin, and must thus be checked against next_epoch.
      • If next_epoch is not known -> undecided: without the next L1 origin we cannot yet determine if time invariant could have been kept.
      • If batch.timestamp >= next_epoch.time -> drop: the batch could have adopted the next L1 origin without breaking the L2 time >= L1 time invariant.
  • len(batch.transactions) > 0: -> drop: when exceeding the sequencer time drift, never allow the sequencer to include transactions.
• batch.transactions: drop if the batch.transactions list contains a transaction that is invalid or derived by other means exclusively:
  • any transaction that is empty (zero length byte string)
  • any deposited transactions (identified by the transaction type prefix byte)

If no batch can be accept-ed, and the stage has completed buffering of all batches that can fully be read from the L1 block at height epoch.number + sequence_window_size, and the next_epoch is available, then an empty batch can be derived with the following properties:

• parent_hash = safe_l2_head.hash
• timestamp = next_timestamp
• transactions is empty, i.e. no sequencer transactions. Deposited transactions may be added in the next stage.
• If next_timestamp < next_epoch.time: the current L1 origin is repeated, to preserve the L2 time invariant.
  • epoch_num = epoch.number
  • epoch_hash = epoch.hash
• If the batch is the first batch of the epoch, that epoch is used instead of advancing the epoch to ensure that there is at least one L2 block per epoch.
  • epoch_num = epoch.number
  • epoch_hash = epoch.hash
• Otherwise,
  • epoch_num = next_epoch.number
  • epoch_hash = next_epoch.hash

Payload Attributes Derivation

In the Payload Attributes Derivation stage, we convert the batches we get from the previous stage into instances of the PayloadAttributes structure. Such a structure encodes the transactions that need to figure into a block, as well as other block inputs (timestamp, fee recipient, etc). Payload attributes derivation is detailed in the section Deriving Payload Attributes section below.

This stage maintains its own copy of the system configuration, independent of the L1 retrieval stage. The system configuration is updated with L1 log events whenever the L1 epoch referenced by the batch input changes.

Engine Queue

In the Engine Queue stage, the previously derived PayloadAttributes structures are buffered and sent to the execution engine to be executed and converted into a proper L2 block.

The stage maintains references to three L2 blocks:

• The finalized L2 head: everything up to and including this block can be fully derived from the finalized (i.e. canonical and forever irreversible) part of the L1 chain.
• The safe L2 head: everything up to and including this block can be fully derived from the currently canonical L1 chain.
• The unsafe L2 head: blocks between the safe and unsafe heads are unsafe blocks that have not been derived from L1. These blocks either come from sequencing (in sequencer mode) or from unsafe sync to the sequencer (in validator mode). This is also known as the "latest" head.

Additionally, it buffers a short history of references to recently processed safe L2 blocks, along with references from which L1 blocks each was derived. This history does not have to be complete, but enables later L1 finality signals to be translated into L2 finality.

Engine API usage

To interact with the engine, the execution engine API is used, with the following JSON-RPC methods:

Bedrock, Canyon, Delta: API Usage

• engine_forkchoiceUpdatedV2 — updates the forkchoice (i.e. the chain head) to headBlockHash if different, and instructs the engine to start building an execution payload if the payload attributes parameter is not null.
• engine_getPayloadV2 — retrieves a previously requested execution payload build.
• engine_newPayloadV2 — executes an execution payload to create a block.

Ecotone: API Usage

• engine_forkchoiceUpdatedV3 — updates the forkchoice (i.e. the chain head) to headBlockHash if different, and instructs the engine to start building an execution payload if the payload attributes parameter is not null.
• engine_getPayloadV3 — retrieves a previously requested execution payload build.
• engine_newPayload
  • engine_newPayloadV2 — executes a Bedrock/Canyon/Delta execution payload to create a block.
  • engine_newPayloadV3 — executes an Ecotone execution payload to create a block.

The current version of op-node uses the v3 Engine API RPC methods as well as engine_newPayloadV2, due to engine_newPayloadV3 only supporting Ecotone execution payloads. Both engine_forkchoiceUpdatedV3 and engine_getPayloadV3 are backwards compatible with Bedrock, Canyon & Delta payloads.

Prior versions of op-node used v2 and v1 methods.

The execution payload is an object of type ExecutionPayloadV3.

The ExecutionPayload has the following requirements:

• Bedrock
  • The withdrawals field MUST be nil
  • The blob gas used field MUST be nil
  • The blob gas limit field MUST be nil
• Canyon, Delta
  • The withdrawals field MUST be non-nil
  • The withdrawals field MUST be an empty list
  • The blob gas used field MUST be nil
  • The blob gas limit field MUST be nil
• Ecotone
  • The withdrawals field MUST be non-nil
  • The withdrawals field MUST be an empty list
  • The blob gas used field MUST be 0
  • The blob gas limit field MUST be 0

Forkchoice synchronization

If there are any forkchoice updates to be applied, before additional inputs are derived or processed, then these are applied to the engine first.

This synchronization may happen when:

• A L1 finality signal finalizes one or more L2 blocks: updating the "finalized" L2 block.
• A successful consolidation of unsafe L2 blocks: updating the "safe" L2 block.
• The first thing after a derivation pipeline reset, to ensure a consistent execution engine forkchoice state.

The new forkchoice state is applied by calling fork choice updated on the engine API. On forkchoice-state validity errors the derivation pipeline must be reset to recover to consistent state.

L1-consolidation: payload attributes matching

If the unsafe head is ahead of the safe head, then consolidation is attempted, verifying that existing unsafe L2 chain matches the derived L2 inputs as derived from the canonical L1 data.

During consolidation, we consider the oldest unsafe L2 block, i.e. the unsafe L2 block directly after the safe head. If the payload attributes match this oldest unsafe L2 block, then that block can be considered "safe" and becomes the new safe head.

The following fields of the derived L2 payload attributes are checked for equality with the L2 block:

• Bedrock, Canyon, Delta, Ecotone Blocks
  • parent_hash
  • timestamp
  • randao
  • fee_recipient
  • transactions_list (first length, then equality of each of the encoded transactions, including deposits)
  • gas_limit
• Canyon, Delta, Ecotone Blocks
  • withdrawals (first presence, then length, then equality of each of the encoded withdrawals)
• Ecotone Blocks
  • parent_beacon_block_root

If consolidation succeeds, the forkchoice change will synchronize as described in the section above.

If consolidation fails, the L2 payload attributes will be processed immediately as described in the section below. The payload attributes are chosen in favor of the previous unsafe L2 block, creating an L2 chain reorg on top of the current safe block. Immediately processing the new alternative attributes enables execution engines like go-ethereum to enact the change, as linear rewinds of the tip of the chain may not be supported.

L1-sync: payload attributes processing

If the safe and unsafe L2 heads are identical (whether because of failed consolidation or not), we send the L2 payload attributes to the execution engine to be constructed into a proper L2 block. This L2 block will then become both the new L2 safe and unsafe head.

If a payload attributes created from a batch cannot be inserted into the chain because of a validation error (i.e. there was an invalid transaction or state transition in the block) the batch should be dropped & the safe head should not be advanced. The engine queue will attempt to use the next batch for that timestamp from the batch queue. If no valid batch is found, the rollup node will create a deposit only batch which should always pass validation because deposits are always valid.

Interaction with the execution engine via the execution engine API is detailed in the Communication with the Execution Engine section.

The payload attributes are then processed with a sequence of:

• Engine: Fork choice updated with current forkchoice state of the stage, and the attributes to start block building.
  • Non-deterministic sources, like the tx-pool, must be disabled to reconstruct the expected block.
• Engine: Get Payload to retrieve the payload, by the payload-ID in the result of the previous step.
• Engine: New Payload to import the new payload into the execution engine.
• Engine: Fork Choice Updated to make the new payload canonical, now with a change of both safe and unsafe fields to refer to the payload, and no payload attributes.

Engine API Error handling:

• On RPC-type errors the payload attributes processing should be re-attempted in a future step.
• On payload processing errors the attributes must be dropped, and the forkchoice state must be left unchanged.
  • Eventually the derivation pipeline will produce alternative payload attributes, with or without batches.
  • If the payload attributes only contained deposits, then it is a critical derivation error if these are invalid.
• On forkchoice-state validity errors the derivation pipeline must be reset to recover to consistent state.

Processing unsafe payload attributes

If no forkchoice updates or L1 data remain to be processed, and if the next possible L2 block is already available through an unsafe source such as the sequencer publishing it via the p2p network, then it is optimistically processed as an "unsafe" block. This reduces later derivation work to just consolidation with L1 in the happy case, and enables the user to see the head of the L2 chain faster than the L1 may confirm the L2 batches.

To process unsafe payloads, the payload must:

• Have a block number higher than the current safe L2 head.
  • The safe L2 head may only be reorged out due to L1 reorgs.
• Have a parent blockhash that matches the current unsafe L2 head.
  • This prevents the execution engine individually syncing a larger gap in the unsafe L2 chain.
  • This prevents unsafe L2 blocks from reorging other previously validated L2 blocks.
  • This check may change in the future versions to adopt e.g. the L1 snap-sync protocol.

The payload is then processed with a sequence of:

• Bedrock/Canyon/Delta Payloads
  • engine_newPayloadV2: process the payload. It does not become canonical yet.
  • engine_forkchoiceUpdatedV2: make the payload the canonical unsafe L2 head, and keep the safe/finalized L2 heads.
• Ecotone Payloads
  • engine_newPayloadV3: process the payload. It does not become canonical yet.
  • engine_forkchoiceUpdatedV3: make the payload the canonical unsafe L2 head, and keep the safe/finalized L2 heads.

Engine API Error handling:

• On RPC-type errors the payload processing should be re-attempted in a future step.
• On payload processing errors the payload must be dropped, and not be marked as canonical.
• On forkchoice-state validity errors the derivation pipeline must be reset to recover to consistent state.

Resetting the Pipeline

It is possible to reset the pipeline, for instance if we detect an L1 reorg (reorganization). This enables the rollup node to handle L1 chain reorg events.

Resetting will recover the pipeline into a state that produces the same outputs as a full L2 derivation process, but starting from an existing L2 chain that is traversed back just enough to reconcile with the current L1 chain.

Note that this algorithm covers several important use-cases:

• Initialize the pipeline without starting from 0, e.g. when the rollup node restarts with an existing engine instance.
• Recover the pipeline if it becomes inconsistent with the execution engine chain, e.g. when the engine syncs/changes.
• Recover the pipeline when the L1 chain reorganizes, e.g. a late L1 block is orphaned, or a larger attestation failure.
• Initialize the pipeline to derive a disputed L2 block with prior L1 and L2 history inside a fault-proof program.

Handling these cases also means a node can be configured to eagerly sync L1 data with 0 confirmations, as it can undo the changes if the L1 later does recognize the data as canonical, enabling safe low-latency usage.

The Engine Queue is first reset, to determine the L1 and L2 starting points to continue derivation from. After this, the other stages are reset independent of each other.

Finding the sync starting point

To find the starting point, there are several steps, relative to the head of the chain traversing back:

1. Find the current L2 forkchoice state
   • If no finalized block can be found, start at the Bedrock genesis block.
   • If no safe block can be found, fallback to the finalized block.
   • The unsafe block should always be available and consistent with the above (it may not be in rare engine-corruption recovery cases, this is being reviewed).
2. Find the first L2 block with plausible L1 reference to be the new unsafe starting point, starting from previous unsafe, back to finalized and no further.
   • Plausible iff: the L1 origin of the L2 block is known and canonical, or unknown and has a block-number ahead of L1.
3. Find the first L2 block with an L1 reference older than the sequencing window, to be the new safe starting point, starting at the above plausible unsafe head, back to finalized and no further.
   • If at any point the L1 origin is known but not canonical, the unsafe head is revised to parent of the current.
   • The highest L2 block with known canonical L1 origin is remembered as highest.
   • If at any point the L1 origin in the block is corrupt w.r.t. derivation rules, then error. Corruption includes:
     • Inconsistent L1 origin block number or parent-hash with parent L1 origin
     • Inconsistent L1 sequence number (always changes to 0 for a L1 origin change, or increments by 1 if not)
   • If the L1 origin of the L2 block n is older than the L1 origin of highest by more than a sequence window, and n.sequence_number == 0, then the parent L2 block of n will be the safe starting point.
4. The finalized L2 block persists as the finalized starting point.
5. Find the first L2 block with an L1 reference older than the channel-timeout
   • The L1 origin referenced by this block which we call l2base will be the base for the L2 pipeline derivation: By starting here, the stages can buffer any necessary data, while dropping incomplete derivation outputs until L1 traversal has caught up with the actual L2 safe head.

While traversing back the L2 chain, an implementation may sanity-check that the starting point is never set too far back compared to the existing forkchoice state, to avoid an intensive reorg because of misconfiguration.

Implementers note: step 1-4 are known as FindL2Heads. Step 5 is currently part of the Engine Queue reset. This may change to isolate the starting-point search from the bare reset logic.

Resetting derivation stages

1. L1 Traversal: start at L1 base as first block to be pulled by next stage.
2. L1 Retrieval: empty previous data, and fetch the base L1 data, or defer the fetching work to a later pipeline step.
3. Frame Queue: empty the queue.
4. Channel Bank: empty the channel bank.
5. Channel Reader: reset any batch decoding state.
6. Batch Queue: empty the batch queue, use base as initial L1 point of reference.
7. Payload Attributes Derivation: empty any batch/attributes state.
8. Engine Queue:
   • Initialize L2 forkchoice state with syncing start point state. (finalized/safe/unsafe)
   • Initialize the L1 point of reference of the stage to base.
   • Require a forkchoice update as first task
   • Reset any finality data

Where necessary, stages starting at base can initialize their system-config from data encoded in the l2base block.

About reorgs Post-Merge

Note that post-merge, the depth of reorgs will be bounded by the L1 finality delay (2 L1 beacon epochs, or approximately 13 minutes, unless more than 1/3 of the network consistently disagrees). New L1 blocks may be finalized every L1 beacon epoch (approximately 6.4 minutes), and depending on these finality-signals and batch-inclusion, the derived L2 chain will become irreversible as well.

Note that this form of finalization only affects inputs, and nodes can then subjectively say the chain is irreversible, by reproducing the chain from these irreversible inputs and the set protocol rules and parameters.

This is however completely unrelated to the outputs posted on L1, which require a form of proof like a fault-proof or zk-proof to finalize. Optimistic-rollup outputs like withdrawals on L1 are only labeled "finalized" after passing a week without dispute (fault proof challenge window), a name-collision with the proof-of-stake finalization.

Deriving Payload Attributes

For every L2 block derived from L1 data, we need to build payload attributes, represented by an expanded version of the PayloadAttributesV2 object, which includes additional transactions and noTxPool fields.

This process happens during the payloads-attributes queue ran by a verifier node, as well as during block-production ran by a sequencer node (the sequencer may enable the tx-pool usage if the transactions are batch-submitted).

Deriving the Transaction List

For each L2 block to be created by the sequencer, we start from a sequencer batch matching the target L2 block number. This could potentially be an empty auto-generated batch, if the L1 chain did not include a batch for the target L2 block number. Remember that the batch includes a sequencing epoch number, an L2 timestamp, and a transaction list.

This block is part of a sequencing epoch, whose number matches that of an L1 block (its L1 origin). This L1 block is used to derive L1 attributes and (for the first L2 block in the epoch) user deposits.

Therefore, a PayloadAttributesV2 object must include the following transactions:

• one or more deposited transactions, of two kinds:
  • a single L1 attributes deposited transaction, derived from the L1 origin.
  • for the first L2 block in the epoch, zero or more user-deposited transactions, derived from the receipts of the L1 origin.
• zero or more network upgrade automation transactions: special transactions to perform network upgrades.
• zero or more sequenced transactions: regular transactions signed by L2 users, included in the sequencer batch.

Transactions must appear in this order in the payload attributes.

The L1 attributes are read from the L1 block header, while deposits are read from the L1 block's receipts. Refer to the deposit contract specification for details on how deposits are encoded as log entries.

Network upgrade automation transactions

Some network upgrades require automated contract changes or deployments at specific blocks. To automate these, without adding persistent changes to the execution-layer, special transactions may be inserted as part of the derivation process.

Ecotone

The Ecotone hardfork activation block contains the following transactions, in this order:

• L1 Attributes Transaction, using the pre-Ecotone setL1BlockValues
• User deposits from L1
• Network Upgrade Transactions
  • L1Block deployment
  • GasPriceOracle deployment
  • Update L1Block Proxy ERC-1967 Implementation Slot
  • Update GasPriceOracle Proxy ERC-1967 Implementation Slot
  • GasPriceOracle Enable Ecotone
  • Beacon block roots contract deployment (EIP-4788)

To not modify or interrupt the system behavior around gas computation, this block will not include any sequenced transactions by setting noTxPool: true.

L1Block Deployment

The L1Block contract is upgraded to process the new Ecotone L1-data-fee parameters and L1 blob base-fee.

A deposit transaction is derived with the following attributes:

• from: 0x4210000000000000000000000000000000000000
• to: null
• mint: 0
• value: 0
• gasLimit: 375,000
• data: 0x60806040523480156100105... (full bytecode)
• sourceHash: 0x877a6077205782ea15a6dc8699fa5ebcec5e0f4389f09cb8eda09488231346f8, computed with the "Upgrade-deposited" type, with `intent = "Ecotone: L1 Block Deployment"

This results in the Ecotone L1Block contract being deployed to 0x07dbe8500fc591d1852B76feE44d5a05e13097Ff, to verify:

cast compute-address --nonce=0 0x4210000000000000000000000000000000000000
Computed Address: 0x07dbe8500fc591d1852B76feE44d5a05e13097Ff

Verify sourceHash:

cast keccak $(cast concat-hex 0x0000000000000000000000000000000000000000000000000000000000000002 $(cast keccak "Ecotone: L1 Block Deployment"))
# 0x877a6077205782ea15a6dc8699fa5ebcec5e0f4389f09cb8eda09488231346f8

Verify data:

git checkout 5996d0bc1a4721f2169ba4366a014532f31ea932
pnpm clean && pnpm install && pnpm build
jq -r ".bytecode.object" packages/contracts-bedrock/forge-artifacts/L1Block.sol/L1Block.json

This transaction MUST deploy a contract with the following code hash 0xc88a313aa75dc4fbf0b6850d9f9ae41e04243b7008cf3eadb29256d4a71c1dfd.

GasPriceOracle Deployment

The GasPriceOracle contract is upgraded to support the new Ecotone L1-data-fee parameters. Post fork this contract will use the blob base fee to compute the gas price for L1-data-fee transactions.

A deposit transaction is derived with the following attributes:

• from: 0x4210000000000000000000000000000000000001
• to: null,
• mint: 0
• value: 0
• gasLimit: 1,000,000
• data: 0x60806040523480156100... (full bytecode)
• sourceHash: 0xa312b4510adf943510f05fcc8f15f86995a5066bd83ce11384688ae20e6ecf42 computed with the "Upgrade-deposited" type, with `intent = "Ecotone: Gas Price Oracle Deployment"

This results in the Ecotone GasPriceOracle contract being deployed to 0xb528D11cC114E026F138fE568744c6D45ce6Da7A, to verify:

cast compute-address --nonce=0 0x4210000000000000000000000000000000000001
Computed Address: 0xb528D11cC114E026F138fE568744c6D45ce6Da7A

Verify sourceHash:

❯ cast keccak $(cast concat-hex 0x0000000000000000000000000000000000000000000000000000000000000002 $(cast keccak "Ecotone: Gas Price Oracle Deployment"))
# 0xa312b4510adf943510f05fcc8f15f86995a5066bd83ce11384688ae20e6ecf42

Verify data:

git checkout 5996d0bc1a4721f2169ba4366a014532f31ea932
pnpm clean && pnpm install && pnpm build
jq -r ".bytecode.object" packages/contracts-bedrock/forge-artifacts/GasPriceOracle.sol/GasPriceOracle.json

This transaction MUST deploy a contract with the following code hash 0x8b71360ea773b4cfaf1ae6d2bd15464a4e1e2e360f786e475f63aeaed8da0ae5.

L1Block Proxy Update

This transaction updates the L1Block Proxy ERC-1967 implementation slot to point to the new L1Block deployment.

A deposit transaction is derived with the following attributes:

• from: 0x0000000000000000000000000000000000000000
• to: 0x4200000000000000000000000000000000000015 (L1Block Proxy)
• mint: 0
• value: 0
• gasLimit: 50,000
• data: 0x3659cfe600000000000000000000000007dbe8500fc591d1852b76fee44d5a05e13097ff
• sourceHash: 0x18acb38c5ff1c238a7460ebc1b421fa49ec4874bdf1e0a530d234104e5e67dbc computed with the "Upgrade-deposited" type, with `intent = "Ecotone: L1 Block Proxy Update"

Verify data:

cast concat-hex $(cast sig "upgradeTo(address)") $(cast abi-encode "upgradeTo(address)" 0x07dbe8500fc591d1852B76feE44d5a05e13097Ff)
0x3659cfe600000000000000000000000007dbe8500fc591d1852b76fee44d5a05e13097ff

Verify sourceHash:

cast keccak $(cast concat-hex 0x0000000000000000000000000000000000000000000000000000000000000002 $(cast keccak "Ecotone: L1 Block Proxy Update"))
# 0x18acb38c5ff1c238a7460ebc1b421fa49ec4874bdf1e0a530d234104e5e67dbc

GasPriceOracle Proxy Update

This transaction updates the GasPriceOracle Proxy ERC-1967 implementation slot to point to the new GasPriceOracle deployment.

A deposit transaction is derived with the following attributes:

• from: 0x0000000000000000000000000000000000000000
• to: 0x420000000000000000000000000000000000000F (Gas Price Oracle Proxy)
• mint: 0
• value: 0
• gasLimit: 50,000
• data: 0x3659cfe6000000000000000000000000b528d11cc114e026f138fe568744c6d45ce6da7a
• sourceHash: 0xee4f9385eceef498af0be7ec5862229f426dec41c8d42397c7257a5117d9230a computed with the "Upgrade-deposited" type, with intent = "Ecotone: Gas Price Oracle Proxy Update"

Verify data:

cast concat-hex $(cast sig "upgradeTo(address)") $(cast abi-encode "upgradeTo(address)" 0xb528D11cC114E026F138fE568744c6D45ce6Da7A)
0x3659cfe6000000000000000000000000b528d11cc114e026f138fe568744c6d45ce6da7a

Verify sourceHash:

cast keccak $(cast concat-hex 0x0000000000000000000000000000000000000000000000000000000000000002 $(cast keccak "Ecotone: Gas Price Oracle Proxy Update"))
# 0xee4f9385eceef498af0be7ec5862229f426dec41c8d42397c7257a5117d9230a

GasPriceOracle Enable Ecotone

This transaction informs the GasPriceOracle to start using the Ecotone gas calculation formula.

A deposit transaction is derived with the following attributes:

• from: 0xDeaDDEaDDeAdDeAdDEAdDEaddeAddEAdDEAd0001 (Depositer Account)
• to: 0x420000000000000000000000000000000000000F (Gas Price Oracle Proxy)
• mint: 0
• value: 0
• gasLimit: 80,000
• data: 0x22b90ab3
• sourceHash: 0x0c1cb38e99dbc9cbfab3bb80863380b0905290b37eb3d6ab18dc01c1f3e75f93, computed with the "Upgrade-deposited" type, with `intent = "Ecotone: Gas Price Oracle Set Ecotone"

Verify data:

cast sig "setEcotone()"
0x22b90ab3

Verify sourceHash:

cast keccak $(cast concat-hex 0x0000000000000000000000000000000000000000000000000000000000000002 $(cast keccak "Ecotone: Gas Price Oracle Set Ecotone"))
# 0x0c1cb38e99dbc9cbfab3bb80863380b0905290b37eb3d6ab18dc01c1f3e75f93

Beacon block roots contract deployment (EIP-4788)

EIP-4788 introduces a "Beacon block roots" contract, that processes and exposes the beacon-block-root values. at address BEACON_ROOTS_ADDRESS = 0x000F3df6D732807Ef1319fB7B8bB8522d0Beac02.

For deployment, EIP-4788 defines a pre-EIP-155 legacy transaction, sent from a key that is derived such that the transaction signature validity is bound to message-hash, which is bound to the input-data, containing the init-code.

However, this type of transaction requires manual deployment and gas-payments. And since the processing is an integral part of the chain processing, and has to be repeated for every OP-Stack chain, the deployment is approached differently here.

Some chains may already have a user-submitted instance of the EIP-4788 transaction. This is cryptographically guaranteed to be correct, but may result in the upgrade transaction deploying a second contract, with the next nonce. The result of this deployment can be ignored.

A Deposit transaction is derived with the following attributes:

• from: 0x0B799C86a49DEeb90402691F1041aa3AF2d3C875, as specified in the EIP.
• to: null
• mint: 0
• value: 0
• gasLimit: 0x3d090, as specified in the EIP.
• isCreation: true
• data: 0x60618060095f395ff33373fffffffffffffffffffffffffffffffffffffffe14604d57602036146024575f5ffd5b5f35801560495762001fff810690815414603c575f5ffd5b62001fff01545f5260205ff35b5f5ffd5b62001fff42064281555f359062001fff015500
• isSystemTx: false, as per the Regolith upgrade, even the system-generated transactions spend gas.
• sourceHash: 0x69b763c48478b9dc2f65ada09b3d92133ec592ea715ec65ad6e7f3dc519dc00c, computed with the "Upgrade-deposited" type, with intent = "Ecotone: beacon block roots contract deployment"

The contract address upon deployment is computed as rlp([sender, nonce]), which will equal:

• BEACON_ROOTS_ADDRESS if deployed
• a different address (0xE3aE1Ae551eeEda337c0BfF6C4c7cbA98dce353B) if nonce = 1: when a user already submitted the EIP transaction before the upgrade.

Verify BEACON_ROOTS_ADDRESS:

cast compute-address --nonce=0 0x0B799C86a49DEeb90402691F1041aa3AF2d3C875
# Computed Address: 0x000F3df6D732807Ef1319fB7B8bB8522d0Beac02

Verify sourceHash:

cast keccak $(cast concat-hex 0x0000000000000000000000000000000000000000000000000000000000000002 $(cast keccak "Ecotone: beacon block roots contract deployment"))
# 0x69b763c48478b9dc2f65ada09b3d92133ec592ea715ec65ad6e7f3dc519dc00c

Building Individual Payload Attributes

After deriving the transactions list, the rollup node constructs a PayloadAttributesV2 as follows:

• timestamp is set to the batch's timestamp.
• random is set to the prev_randao L1 block attribute.
• suggestedFeeRecipient is set to the Sequencer Fee Vault address. See Fee Vaults specification.
• transactions is the array of the derived transactions: deposited transactions and sequenced transactions, all encoded with EIP-2718.
• noTxPool is set to true, to use the exact above transactions list when constructing the block.
• gasLimit is set to the current gasLimit value in the system configuration of this payload.
• withdrawals is set to nil prior to Canyon and an empty array after Canyon

Batch Submitter

Table of Contents

• Overview

Overview

The batch submitter, also referred to as the batcher, is the entity submitting the L2 sequencer data to L1, to make it available for verifiers.

The format of the data transactions is defined in the derivation spec: the data is constructed from L2 blocks in the reverse order as it is derived from data into L2 blocks.

The timing, operation and transaction signing is implementation-specific: any data can be submitted at any time, but only the data that matches the derivation spec rules will be valid from the verifier perspective.

The most minimal batcher implementation can be defined as a loop of the following operations:

1. See if the unsafe L2 block number is past the safe block number: unsafe data needs to be submitted.
2. Iterate over all unsafe L2 blocks, skip any that were previously submitted.
3. Open a channel, buffer all the L2 block data to be submitted, while applying the encoding and compression as defined in the derivation spec.
4. Pull frames from the channel to fill data transactions with, until the channel is empty.
5. Submit the data transactions to L1

The L2 view of safe/unsafe does not instantly update after data is submitted, nor when it gets confirmed on L1, so special care may have to be taken to not duplicate data submissions.

Fault Proof

Table of Contents

• Overview
• Pre-image Oracle
  • Pre-image key types
    • Type 0: Zero key
    • Type 1: Local key
    • Type 2: Global keccak256 key
    • Type 3: Global generic key
    • Type 4: Global SHA2-256 key
    • Type 5: Global EIP-4844 Point-evaluation key
    • Type 6: Global Precompile key
    • Type 7-128: reserved range
    • Type 129-255: application usage
  • Bootstrapping
  • Hinting
  • Pre-image communication
• Fault Proof Program
  • Prologue
  • Main content
  • Epilogue
  • Pre-image hinting routes
    • l1-block-header <blockhash>
    • l1-transactions <blockhash>
    • l1-receipts <blockhash>
    • l1-precompile <precompile ++ inputbytes>
    • l2-block-header <blockhash>
    • l2-transactions <blockhash>
    • l2-code <codehash>
    • l2-state-node <nodehash>
    • l2-output <outputroot>
    • l2-payload-witness <payload_attributes>
    • l2-account-proof <blockhash_and_address>
  • Precompile Accelerators
• Fault Proof VM
• Fault Proof Interactive Dispute Game

Overview

A fault proof, also known as fraud proof or interactive game, consists of 3 components:

• Program: given a commitment to all rollup inputs (L1 data) and the dispute, verify the dispute statelessly.
• VM: given a stateless program and its inputs, trace any instruction step, and prove it on L1.
• Interactive Dispute Game: bisect a dispute down to a single instruction, and resolve the base-case using the VM.

Each of these 3 components may have different implementations, which can be combined into different proof stacks, and contribute to proof diversity when resolving a dispute.

"Stateless execution" of the program, and its individual instructions, refers to reproducing the exact same computation by authenticating the inputs with a Pre-image Oracle.

Pre-image Oracle

The pre-image oracle is the only form of communication between the Program (in the Client role) and the VM (in the Server role).

The program uses the pre-image oracle to query any input data that is understood to be available to the user:

• The initial inputs to bootstrap the program. See Bootstrapping.
• External data not already part of the program code. See Pre-image hinting routes.

The communication happens over a simple request-response wire protocol, see Pre-image communication.

Pre-image key types

Pre-images are identified by a bytes32 type-prefixed key:

• The first byte identifies the type of request.
• The remaining 31 bytes identify the pre-image key.

Type 0: Zero key

The zero prefix is illegal. This ensures all pre-image keys are non-zero, enabling storage lookup optimizations and avoiding easy mistakes with invalid zeroed keys in the EVM.

Type 1: Local key

Information specific to the dispute: the remainder of the key may be an index, a string, a hash, etc. Only the contract(s) managing this dispute instance may serve the value for this key: it is localized and context-dependent.

This type of key is used for program bootstrapping, to identify the initial input arguments by index or name.

Type 2: Global keccak256 key

This type of key uses a global pre-image store contract, and is fully context-independent and permissionless. I.e. every key must have a single unique value, regardless of chain history or time.

Using a global store reduces duplicate pre-image registration work, and avoids unnecessary contract redeployments per dispute.

This global store contract should be non-upgradeable.

Since keccak256 is a safe 32-byte hash input, the first byte is overwritten with a 2 to derive the key, while keeping the rest of the key "readable" (matching the original hash).

Type 3: Global generic key

Reserved. This scheme allows for unlimited application-layer pre-image types without fault-proof VM redeployments.

This is a generic version of a global key store: key = 0x03 ++ keccak256(x, sender)[1:], where:

• x is a bytes32, which can be a hash of an arbitrary-length type of cryptographically secure commitment.
• sender is a bytes32 identifying the pre-image inserter address (left-padded to 32 bytes)

This global store contract should be non-upgradeable.

The global contract is permissionless: users can standardize around external contracts that verify pre-images (i.e. a specific sender will always be trusted for a specific kind of pre-image). The external contract verifies the pre-image before inserting it into the global store for usage by all fault proof VMs without requiring the VM or global store contract to be changed.

Users may standardize around upgradeable external pre-image contracts, in case the implementation of the verification of the pre-image is expected to change.

The store update function is update(x bytes32, offset uint64, span uint8, value bytes32):

• x is the bytes32 x that the pre-image key is computed with.
• Only part of the pre-image, starting at offset, and up to (incl.) 32 bytes span can be inserted at a time.
• Pre-images may have an undefined length (e.g. a stream), we only need to know how many bytes of value are usable.
• The key and offset will be hashed together to uniquely store the value and span, for later pre-image serving.

This enables fault proof programs to adopt any new pre-image schemes without VM update or contract redeployment.

It is up to the user to index the special pre-image values by this key scheme, as there is no way to revert it to the original commitment without knowing said commitment or value.

Type 4: Global SHA2-256 key

A SHA-256 pre-image.

Key: the SHA-256 hash, with the first byte overwritten with the type byte: 4 ++ sha256(data)[1:].

Type 5: Global EIP-4844 Point-evaluation key

An EIP-4844 point-evaluation. In an EIP-4844 blob, 4096 field elements represent the blob data.

It verifies p(z) = y given commitment that corresponds to the polynomial p(x) and a KZG proof. The value y is the pre-image. The value z is part of the key; the index of the point within the blob. The commitment is part of the key.

Each element is proven with a point-evaluation.

Key: 5 ++ keccak256(commitment ++ z)[1:], where:

• 5 is the type byte
• ++ is concatenation
• commitment is a bytes48, representing the KZG commitment.
• z is a big-endian uint256

Type 6: Global Precompile key

A precompile result. It maps directly to precompiles on Ethereum.

This preimage key can be used to avoid running expensive precompile operations in the program.

Key: 6 ++ keccak256(precompile ++ input)[1:], where:

• 6 is the type byte
• ++ is concatenation
• precompile is the 20-byte address of the precompile contract
• input is the input to the precompile contract

The result is identical to that of a call to the precompile contract, prefixed with a revert indicator:

• reverted ++ precompile_result.

reverted is a 1-byte indicator with a 0 value if the precompile reverts for the given input, otherwise it's 1.

Type 7-128: reserved range

Range start and end both inclusive.

This range of key types is reserved for future usage by the core protocol. E.g. version changes, contract migrations, chain-data, additional core features, etc.

128 specifically (1000 0000 in binary) is reserved for key-type length-extension (reducing the content part to 30 or less key bytes), if the need arises.

Type 129-255: application usage

This range of key types may be used by forks or customized versions of the fault proof protocol.

Bootstrapping

Initial inputs are deterministic, but not necessarily singular or global: there may be multiple different disputes at the same time, each with its own disputed claims and L1 context.

To bootstrap, the program requests the initial inputs from the VM, using pre-image key type 1.

The VM is aware of the external context, and maps requested pre-image keys based on their type, i.e. a local lookup for type 1, or global one for 2, and optionally support other key-types.

Hinting

There is one more form of optional communication between client and server: pre-image hinting. Hinting is optional, and is a no-op in a L1 VM implementation.

The hint itself comes at very low cost onchain: the hint can be a single write sys-call, which is instant as the memory to write as hint does not actually need to be loaded as part of the onchain proof.

Hinting allows the program, when generating a proof offchain, to instruct the VM what data it is interested in.

The VM can choose to execute the requested hint at any time: either locally (for standard requests), or in a modular form by redirecting the hint to tooling that may come with the VM program.

Hints do not have to be executed directly: they may first just be logged to show the intents of the program, and the latest hint may be buffered for lazy execution, or dropped entirely when in read-only mode (like onchain).

When the pre-image oracle serves a request, and the request cannot be served from an existing collection of pre-images (e.g. a local pre-image cache) then the VM can execute the hint to retrieve the missing pre-image(s). It is the responsibility of the program to provide sufficient hinting for every pre-image request. Some hints may have to be repeated: the VM only has to execute the last hint when handling a missing pre-image.

Note that hints may produce multiple pre-images: e.g. a hint for an ethereum block with transaction list may prepare pre-images for the header, each of the transactions, and the intermediate merkle-nodes that form the transactions-list Merkle Patricia Trie.

Hinting is implemented with a request-acknowledgement wire-protocol over a blocking two-way stream:

<request> := <length prefix> <hint bytes>

<response> := <ack>

<length prefix> := big-endian uint32  # length of <hint bytes>
<hint bytes> := byte sequence
<ack> := 1-byte zero value

The ack informs the client that the hint has been processed. Servers may respond to hints and pre-image (see below) requests asynchronously as they are on separate streams. To avoid requesting pre-images that are not yet fetched, clients should request the pre-image only after it has observed the hint acknowledgement.

Pre-image communication

Pre-images are communicated with a minimal wire-protocol over a blocking two-way stream. This protocol can be implemented with blocking read/write syscalls.

<request> := <bytes32>  # the type-prefixed pre-image key

<response> := <length prefix> <pre-image bytes>

<length prefix> := big-endian uint64  # length of <pre-image bytes>, note: uint64

The <length prefix> here may be arbitrarily high: the client can stop reading at any time if the required part of the pre-image has been read.

After the client writes new <request> bytes, the server should be prepared to respond with the pre-image starting from offset == 0 upon read calls.

The server may limit read results artificially to only a small amount of bytes at a time, even though the full pre-image is ready: this is expected regular IO protocol, and the client will just have to continue to read the small results at a time, until 0 bytes are read, indicating EOF. This enables the server to serve e.g. at most 32 bytes at a time or align reads with VM memory structure, to limit the amount of VM state that changes per syscall instruction, and thus keep the proof size per instruction bounded.

Fault Proof Program

The Fault Proof Program defines the verification of claims of the state-transition outputs of the L2 rollup as a pure function of L1 data.

The op-program is the reference implementation of the program, based on op-node and op-geth implementations.

The program consists of:

• Prologue: load the inputs, given minimal bootstrapping, with possible test-overrides.
• Main content: process the L2 state-transition, i.e. derive the state changes from the L1 inputs.
• Epilogue: inspect the state changes to verify the claim.

Prologue

The program is bootstrapped with two primary inputs:

• l1_head: the L1 block hash that will be perceived as the tip of the L1 chain, authenticating all prior L1 history.
• dispute: identity of the claim to verify.

Bootstrapping happens through special input requests to the host of the program.

Additionally, there are implied inputs, which are derived from the above primary inputs, but can be overridden for testing purposes:

• l2_head: the L2 block hash that will be perceived as the previously agreed upon tip of the L2 chain, authenticating all prior L2 history.
• Chain configurations: chain configuration may be baked into the program, or determined from attributes of the identified dispute on L1.
  • l1_chain_config: The chain-configuration of the L1 chain (also known as l1_genesis.json)
  • l2_chain_config: The chain-configuration of the L2 chain (also known as l2_genesis.json)
  • rollup_config: The rollup configuration used by the rollup-node (also known as rollup.json)

The implied inputs rely on L1-introspection to load attributes of the dispute through the dispute game interface, in the L1 history up and till the specified l1_head. The dispute may be the claim itself, or a pointer to specific prior claimed data in L1, depending on the dispute game interface.

Implied inputs are loaded in a "prologue" before the actual core state-transition function executes. During testing a simplified prologue that loads the overrides may be used.

Note: only the test-prologues are currently supported, since the dispute game interface is actively changing.

Main content

To verify a claim about L2 state, the program first reproduces the L2 state by applying L1 data to prior agreed L2 history.

This process is also known as the L2 derivation process, and matches the processing in the rollup node and execution-engine.

The difference is that rather than retrieving inputs from an RPC and applying state changes to disk, the inputs are loaded through the pre-image oracle and the changes accumulate in memory.

The derivation executes with two data-sources:

• Interface to read-only L1 chain, backed by the pre-image oracle:
  • The l1_head determines the view over the available L1 data: no later L1 data is available.
  • The implementation of the chain traverses the header-chain from the l1_head down to serve by-number queries.
  • The l1_head is the L1 unsafe head, safe head, and finalized head.
• Interface to L2 engine API
  • Prior L2 chain history is backed by the pre-image oracle, similar to the L1 chain:
    • The initial l2_head determines the view over the initial available L2 history: no later L2 data is available.
    • The implementation of the chain traverses the header-chain from the l2_head down to serve by-number queries.
    • The l2_head is the initial L2 unsafe head, safe head, and finalized head.
  • New L2 chain history accumulates in memory.
    • Although the pre-image oracle can be used to retrieve data by hash if memory is limited, the program should prefer to keep the newly created chain data in memory, to minimize pre-image oracle access.
    • The L2 unsafe head, safe head, and finalized L2 head will potentially change as derivation progresses.
    • L2 state consists of the diff of changes in memory, and any unchanged state nodes accessible through the read-only L2 history view.

See Pre-image routes for specifications of the pre-image oracle backing of these data sources.

Using these data-sources, the derivation pipeline is processed till we hit one of two conditions:

• EOF: when we run out of L1 data, the L2 chain will not change further, and the epilogue can start.
• Eager epilogue condition: depending on the type of claim to verify, if the L2 result is irreversible (i.e. no later L1 inputs can override it), the processing may end early when the result is ready. E.g. when asserting state at a specific L2 block, rather than the very tip of the L2 chain.

Epilogue

While the main-content produces the disputed L2 state already, the epilogue concludes what this means for the disputed claim.

The program produces a binary output to verify the claim, using a standard single-byte Unix exit-code:

• a 0 for success, i.e. the claim is correct.
• a non-zero code for failure, i.e. the claim is incorrect.
  • 1 should be preferred for identifying an incorrect claim.
  • Other non-zero exit codes may indicate runtime failure, e.g. a bug in the program code may resolve in a kind of panic or unexpected error. Safety should be preferred over liveness in this case, and the claim will fail.

To assert the disputed claim, the epilogue, like the main content, can introspect L1 and L2 chain data and post-process it further, to then make a statement about the claim with the final exit code.

A disputed output-root may be disproven by first producing the output-root, and then comparing it:

1. Retrieve the output attributes from the L2 chain view: the state-root, block-hash, withdrawals storage-root.
2. Compute the output-root, as the proposer should compute it.
3. If the output-root matches the claim, exit with code 0. Otherwise, exit with code 1.

Note: the dispute game interface is actively changing, and may require additional claim assertions. the output-root epilogue may be replaced or extended for general L2 message proving.

Pre-image hinting routes

The fault proof program implements hint handling for the VM to use, as well as any program testing outside of VM environment. This can be exposed via a CLI, or alternative inter-process API.

Every instance of <blockhash> in the below routes is 0x-prefixed, lowercase, hex-encoded.

l1-block-header <blockhash>

Requests the host to prepare the L1 block header RLP pre-image of the block <blockhash>.

l1-transactions <blockhash>

Requests the host to prepare the list of transactions of the L1 block with <blockhash>: prepare the RLP pre-images of each of them, including transactions-list MPT nodes.

l1-receipts <blockhash>

Requests the host to prepare the list of receipts of the L1 block with <blockhash>: prepare the RLP pre-images of each of them, including receipts-list MPT nodes.

l1-precompile <precompile ++ inputbytes>

Requests the host to prepare the result of an L1 call to the precompile address given <inputbytes> as the input. The host also prepares a global keccak256 preimage of the hint data <precompile ++ inputbytes>.

l2-block-header <blockhash>

Requests the host to prepare the L2 block header RLP pre-image of the block <blockhash>.

l2-transactions <blockhash>

Requests the host to prepare the list of transactions of the L2 block with <blockhash>: prepare the RLP pre-images of each of them, including transactions-list MPT nodes.

l2-code <codehash>

Requests the host to prepare the L2 smart-contract code with the given <codehash>.

l2-state-node <nodehash>

Requests the host to prepare the L2 MPT node preimage with the given <nodehash>.

l2-output <outputroot>

Requests the host to prepare the L2 Output at the l2 output root <outputroot>. The L2 Output is the preimage of a computed output root.

l2-payload-witness <payload_attributes>

Requests the host to prepare all preimages used in the building of the payload specified by <payload_attributes>. <payload_attributes> is a JSON object with the fields parentBlockHash and payloadAttributes.

l2-account-proof <blockhash_and_address>

Requests the host send account proof for a certain block hash and address. <blockhash_and_address> is hex encoded: 32-byte block hash + 20-byte address.

l2-payload-witness and l2-account-proof hints are preferred over the more granular l2-code and l2-state-node, and they should be sent before the more granular hints to ensure proper handling.

Precompile Accelerators

Precompiles that are too expensive to be executed in a fault-proof VM can be executed more efficiently using the pre-image oracle. This approach ensures that the fault proof program can complete a state transition in a reasonable amount of time.

During program execution, the precompiles are substituted with interactions with pre-image oracle. The program hints the host for a precompile input. Which it the subsequently retrieves the result of the precompile operation using the type 6 global precompile key. All accelerated precompiles must be functionally equivalent to their EVM equivalent.

Fault Proof VM

A fault proof VM implements:

• a smart-contract to verify a single execution-trace step, e.g. a single MIPS instruction.
• a CLI command to generate a proof of a single execution-trace step.
• a CLI command to compute a VM state-root at step N

A fault proof VM relies on a fault proof program to provide an interface for fetching any missing pre-images based on hints.

The VM emulates the program, as prepared for the VM target architecture, and generates the state-root or instruction proof data as requested through the VM CLI.

Refer to the documentation of the fault proof VM for further usage information.

Fault Proof VMs:

• Cannon: big-endian 32-bit MIPS proof, by OP Labs, in active development.
• cannon-rs: Rust implementation of Cannon, by @clabby, in active development.
• Asterisc: little-endian 64-bit RISC-V proof, by @protolambda, in active development.

Fault Proof Interactive Dispute Game

The interactive dispute game allows actors to resolve a dispute with an onchain challenge-response game that bisects to a disagreed block $n\to n+1$ state transition, and then over the execution trace of the VM which models this state transition, bounded with a base-case that proves a single VM trace step.

The game is multi-player: different non-aligned actors may participate when bonded.

Response time is allocated based on the remaining time in the branch of the tree and alignment with the claim. The allocated response time is limited by the dispute-game window, and any additional time necessary based on L1 fee changes when bonds are insufficient.

Note: the timed, bonded, bisection dispute game is in development. Also see fault dispute-game specs for fault dispute game system specifications, And dispute-game-interface specs

Cannon Fault Proof Virtual Machine

Table of Contents

• Overview
• State
  • State Hash
• Memory
  • Heap
• Delay Slots
• Syscalls
• I/O
  • Standard Streams
  • Hint Communication
  • Pre-image Communication
    • Pre-image I/O Alignment
• Exceptions
• Security Model
  • Compiler Correctness
  • Compiler Assumptions

Overview

This is a description of the Cannon Fault Proof Virtual Machine (FPVM). The Cannon FPVM emulates a minimal Linux-based system running on big-endian 32-bit MIPS32 architecture. A lot of its behaviors are copied from Linux/MIPS with a few tweaks made for fault proofs. For the rest of this doc, we refer to the Cannon FPVM as simply the FPVM.

Operationally, the FPVM is a state transition function. This state transition is referred to as a Step, that executes a single instruction. We say the VM is a function $f$, given an input state $S_{pre}$, steps on a single instruction encoded in the state to produce a new state $S_{post}$. $$f(S_{pre})\to S_{post}$$

Thus, the trace of a program executed by the FPVM is an ordered set of VM states.

State

The virtual machine state highlights the effects of running a Fault Proof Program on the VM. It consists of the following fields:

1. memRoot - A bytes32 value representing the merkle root of VM memory.
2. preimageKey - bytes32 value of the last requested pre-image key.
3. preimageOffset - The 32-bit value of the last requested pre-image offset.
4. pc - 32-bit program counter.
5. nextPC - 32-bit next program counter. Note that this value may not always be $pc+4$ when executing a branch/jump delay slot.
6. lo - 32-bit MIPS LO special register.
7. hi - 32-bit MIPS HI special register.
8. heap - 32-bit base address of the most recent memory allocation via mmap.
9. exitCode - 8-bit exit code.
10. exited - 8-bit indicator that the VM has exited.
11. step - 8-byte step counter.
12. registers - General-purpose MIPS32 registers. Each register is a 32-bit value.

The state is represented by packing the above fields, in order, into a 226-byte buffer.

State Hash

The state hash is computed by hashing the 226-byte state buffer with the Keccak256 hash function and then setting the high-order byte to the respective VM status.

The VM status can be derived from the state's exited and exitCode fields.

enum VmStatus {
    Valid = 0,
    Invalid = 1,
    Panic = 2,
    Unfinished = 3,
}

fn vm_status(exit_code: u8, exited: bool) -> u8 {
    if exited {
        match exit_code {
            0 => VmStatus::Valid,
            1 => VmStatus::Invalid,
            _ => VmStatus::Panic,
        }
    } else {
        VmStatus::Unfinished
    }
}

Memory

Memory is represented as a binary merkle tree. The tree has a fixed-depth of 27 levels, with leaf values of 32 bytes each. This spans the full 32-bit address space, where each leaf contains the memory at that part of the tree. The state memRoot represents the merkle root of the tree, reflecting the effects of memory writes. As a result of this memory representation, all memory operations are 4-byte aligned. Memory access doesn't require any privileges. An instruction step can access any memory location as the entire address space is unprotected.

Heap

FPVM state contains a heap that tracks the base address of the most recent memory allocation. Heap pages are bump allocated at the page boundary, per mmap syscall. mmap-ing is purely to satisfy program runtimes that need the memory-pointer result of the syscall to locate free memory. The page size is 4096.

The FPVM has a fixed program break at 0x40000000. However, the FPVM is permitted to extend the heap beyond this limit via mmap syscalls. For simplicity, there are no memory protections against "heap overruns" against other memory segments. Such VM steps are still considered valid state transitions.

Specification of memory mappings is outside the scope of this document as it is irrelevant to the VM state. FPVM implementers may refer to the Linux/MIPS kernel for inspiration.

Delay Slots

The post-state of a step updates the nextPC, indicating the instruction following the pc. However, in the case of where a branch instruction is being stepped, the nextPC post-state is set to the branch target. And the pc post-state set to the branch delay slot as usual.

A VM state transition is invalid whenever the current instruction is a delay slot that is filled with jump or branch type instruction. That is, where $nextPC\ne pc+4$ while stepping on a jump/branch instruction. Otherwise, there would be two consecutive delay slots. While this is considered "undefined" behavior in typical MIPS implementations, FPVM must raise an exception when stepping on such states.

Syscalls

Syscalls work similar to Linux/MIPS, including the syscall calling conventions and general syscall handling behavior. However, the FPVM supports a subset of Linux/MIPS syscalls with slightly different behaviors. The following table list summarizes the supported syscalls and their behaviors.

For all of the above syscalls, an error is indicated by setting the return register ($v0) to 0xFFFFFFFF (-1) and errno ($a3) is set accordingly. The VM must not modify any register other than $v0 and $a3 during syscall handling. For unsupported syscalls, the VM must do nothing except to zero out the syscall return ($v0) and errno ($a3) registers.

Note that the above syscalls have identical syscall numbers and ABIs as Linux/MIPS.

I/O

The VM does not support Linux open(2). However, the VM can read from and write to a predefined set of file descriptors.

Syscalls referencing unknown file descriptors fail with an EBADF errno as done on Linux.

Writing to and reading from standard output, input and error streams have no effect on the FPVM state. FPVM implementations may use them for debugging purposes as long as I/O is stateless.

All I/O operations are restricted to a maximum of 4 bytes per operation. Any read or write syscall request exceeding this limit will be truncated to 4 bytes. Consequently, the return value of read/write syscalls is at most 4, indicating the actual number of bytes read/written.

Standard Streams

Writing to stderr/stdout standard stream always succeeds with the write count input returned, effectively continuing execution without writing work. Reading from stdin has no effect other than to return zero and errno set to 0, signalling that there is no input.

Hint Communication

Hint requests and responses have no effect on the VM state other than setting the $v0 return register to the requested read/write count. VM implementations may utilize hints to setup subsequent pre-image requests.

Pre-image Communication

The preimageKey and preimageOffset state are updated via read/write syscalls to the pre-image read and write file descriptors (see I/O). The preimageKey buffers the stream of bytes written to the pre-image write fd. The preimageKey buffer is shifted to accommodate new bytes written to the end of it. A write also resets the preimageOffset to 0, indicating the intent to read a new pre-image.

When handling pre-image reads, the preimageKey is used to lookup the pre-image data from an Oracle. A max 4-byte chunk of the pre-image at the preimageOffset is read to the specified address. Each read operation increases the preimageOffset by the number of bytes requested (truncated to 4 bytes and subject to alignment constraints).

Pre-image I/O Alignment

As mentioned earlier in memory, all memory operations are 4-byte aligned. Since pre-image I/O occurs on memory, all pre-image I/O operations must strictly adhere to alignment boundaries. This means the start and end of a read/write operation must fall within the same alignment boundary. If an operation were to violate this, the input count of the read/write syscall must be truncated such that the effective address of the last byte read/written matches the input effective address.

The VM must read/write the maximum amount of bytes possible without crossing the input address alignment boundary. For example, the effect of a write request for a 3-byte aligned buffer must be exactly 3 bytes. If the buffer is misaligned, then the VM may write less than 3 bytes depending on the size of the misalignment.

Exceptions

The FPVM may raise an exception rather than output a post-state to signal an invalid state transition. Nominally, the FPVM must raise an exception in at least the following cases:

• Invalid instruction (either via an invalid opcode or an instruction referencing registers outside the general purpose registers).
• Pre-image read at an offset larger than the size of the pre-image.
• Delay slot contains branch/jump instruction types.

VM implementations may raise an exception in other cases that is specific to the implementation. For example, an on-chain FPVM that relies on pre-supplied merkle proofs for memory access may raise an exception if the supplied merkle proof does not match the pre-state memRoot.

Security Model

Compiler Correctness

Cannon is designed to prove the correctness of a particular state transition that emulates a MIPS32 machine. Cannon does not guarantee that the MIPS32 instructions correctly implement the program that the user intends to prove. As a result, Cannon's use as a Fault Proof system inherently depends on some extent on the correctness of the compiler used to generate the MIPS32 instructions over which Cannon operates.

To illustrate this concept, suppose that a user intends to prove simple program input + 1 = output. Suppose then that the user's compiler for this program contains a bug and errantly generates the MIPS instructions for a slightly different program input + 2 = output. Although Cannon would correctly prove the operation of this compiled program, the result proven would differ from the user's intent. Cannon proves the MIPS state transition but makes no assertion about the correctness of the translation between the user's high-level code and the resulting MIPS program.

As a consequence of the above, it is the responsibility of a program developer to develop tests that demonstrate that Cannon is capable of proving their intended program correctly over a large number of possible inputs. Such tests defend against bugs in the user's compiler as well as ways in which the compiler may inadvertently break one of Cannon's Compiler Assumptions. Users of Fault Proof systems are strongly encouraged to utilize multiple proof systems and/or compilers to mitigate the impact of errant behavior in any one toolchain.

Compiler Assumptions

Cannon makes the simplifying assumption that users are utilizing compilers that do not rely on MIPS exception states for standard program behavior. In other words, Cannon generally assumes that the user's compiler generates spec-compliant instructions that would not trigger an exception. Refer to Exceptions for a list of conditions that are explicitly handled.

Certain cases that would typically be asserted by a strict implementation of the MIPS32 specification are not handled by Cannon as follows:

• add, addi, and sub do not trigger an exception on signed integer overflow.
• Instruction encoding validation does not trigger an exception for fields that should be zero.
• Memory instructions do not trigger an exception when addresses are not naturally aligned.

Many compilers, including the Golang compiler, will not generate code that would trigger these conditions under bug-free operation. Given the inherent reliance on Compiler Correctness in applications using Cannon, the tests and defense mechanisms that must necessarily be employed by Cannon users to protect their particular programs against compiler bugs should also suffice to surface bugs that would break these compiler assumptions. Stated simply, Cannon can rely on specific compiler behaviors because users inherently must employ safety nets to guard against compiler bugs.

Stage One Decentralization

This section of the specification contains the system design for stage one decentralization, with a fault-proof system for output proposals and the integration with the OptimismPortal contract, which is the arbiter of withdrawals on L1.

Dispute Game Interface

Table of Contents

• Overview
• Types
• DisputeGameFactory Interface
• DisputeGame Interface

Overview

A dispute game is played between multiple parties when contesting the truthiness of a claim. In the context of an optimistic rollup, claims are made about the state of the layer two network to enable withdrawals to the layer one. A proposer makes a claim about the layer two state such that they can withdraw and a challenger can dispute the validity of the claim. The security of the layer two comes from the ability of fraudulent withdrawals being able to be disputed.

A dispute game interface is defined to allow for multiple implementations of dispute games to exist. If multiple dispute games run in production, it gives a similar security model as having multiple protocol clients, as a bug in a single dispute game will not result in the bug becoming consensus.

Types

For added context, we define a few types that are used in the following snippets.

/// @notice A `Claim` type represents a 32 byte hash or other unique identifier for a claim about
///         a certain piece of information.
type Claim is bytes32;

/// @notice A custom type for a generic hash.
type Hash is bytes32;

/// @notice A dedicated timestamp type.
type Timestamp is uint64;

/// @notice A `GameType` represents the type of game being played.
type GameType is uint32;

/// @notice A `GameId` represents a packed 4 byte game type, a 8 byte timestamp, and a 20 byte address.
/// @dev The packed layout of this type is as follows:
/// ┌───────────┬───────────┐
/// │   Bits    │   Value   │
/// ├───────────┼───────────┤
/// │ [0, 32)   │ Game Type │
/// │ [32, 96)  │ Timestamp │
/// │ [96, 256) │ Address   │
/// └───────────┴───────────┘
type GameId is bytes32;

/// @title GameTypes
/// @notice A library that defines the IDs of games that can be played.
library GameTypes {
    /// @dev A dispute game type the uses the cannon vm.
    GameType internal constant CANNON = GameType.wrap(0);

    /// @dev A dispute game type that performs output bisection and then uses the cannon vm.
    GameType internal constant OUTPUT_CANNON = GameType.wrap(1);

    /// @notice A dispute game type that performs output bisection and then uses an alphabet vm.
    ///         Not intended for production use.
    GameType internal constant OUTPUT_ALPHABET = GameType.wrap(254);

    /// @notice A dispute game type that uses an alphabet vm.
    ///         Not intended for production use.
    GameType internal constant ALPHABET = GameType.wrap(255);
}

/// @notice The current status of the dispute game.
enum GameStatus {
    /// @dev The game is currently in progress, and has not been resolved.
    IN_PROGRESS,
    /// @dev The game has concluded, and the `rootClaim` was challenged successfully.
    CHALLENGER_WINS,
    /// @dev The game has concluded, and the `rootClaim` could not be contested.
    DEFENDER_WINS
}

DisputeGameFactory Interface

The dispute game factory is responsible for creating new DisputeGame contracts given a GameType and a root Claim. Challenger agents listen to the DisputeGameCreated events in order to keep up with on-going disputes in the protocol and participate accordingly.

A clones-with-immutable-args factory (originally by @wighawag, but forked and improved by @Vectorized) is used to create Clones. Each GameType has a corresponding implementation within the factory, and when a new game is created, the factory creates a clone of the GameType's pre-deployed implementation contract.

The rootClaim of created dispute games can either be a claim that the creator agrees or disagrees with. This is an implementation detail that is left up to the IDisputeGame to handle within its resolve function.

When the DisputeGameFactory creates a new DisputeGame contract, it calls initialize() on the clone to set up the game.

/// @title IDisputeGameFactory
/// @notice The interface for a DisputeGameFactory contract.
interface IDisputeGameFactory {
    /// @notice Emitted when a new dispute game is created
    /// @param disputeProxy The address of the dispute game proxy
    /// @param gameType The type of the dispute game proxy's implementation
    /// @param rootClaim The root claim of the dispute game
    event DisputeGameCreated(address indexed disputeProxy, GameType indexed gameType, Claim indexed rootClaim);

    /// @notice Emitted when a new game implementation added to the factory
    /// @param impl The implementation contract for the given `GameType`.
    /// @param gameType The type of the DisputeGame.
    event ImplementationSet(address indexed impl, GameType indexed gameType);

    /// @notice Emitted when a game type's initialization bond is updated
    /// @param gameType The type of the DisputeGame.
    /// @param newBond The new bond (in wei) for initializing the game type.
    event InitBondUpdated(GameType indexed gameType, uint256 indexed newBond);

    /// @notice Information about a dispute game found in a `findLatestGames` search.
    struct GameSearchResult {
        uint256 index;
        GameId metadata;
        Timestamp timestamp;
        Claim rootClaim;
        bytes extraData;
    }

    /// @notice The total number of dispute games created by this factory.
    /// @return gameCount_ The total number of dispute games created by this factory.
    function gameCount() external view returns (uint256 gameCount_);

    /// @notice `games` queries an internal mapping that maps the hash of
    ///         `gameType ++ rootClaim ++ extraData` to the deployed `DisputeGame` clone.
    /// @dev `++` equates to concatenation.
    /// @param _gameType The type of the DisputeGame - used to decide the proxy implementation
    /// @param _rootClaim The root claim of the DisputeGame.
    /// @param _extraData Any extra data that should be provided to the created dispute game.
    /// @return proxy_ The clone of the `DisputeGame` created with the given parameters.
    ///         Returns `address(0)` if nonexistent.
    /// @return timestamp_ The timestamp of the creation of the dispute game.
    function games(
        GameType _gameType,
        Claim _rootClaim,
        bytes calldata _extraData
    )
        external
        view
        returns (IDisputeGame proxy_, Timestamp timestamp_);

    /// @notice `gameAtIndex` returns the dispute game contract address and its creation timestamp
    ///          at the given index. Each created dispute game increments the underlying index.
    /// @param _index The index of the dispute game.
    /// @return gameType_ The type of the DisputeGame - used to decide the proxy implementation.
    /// @return timestamp_ The timestamp of the creation of the dispute game.
    /// @return proxy_ The clone of the `DisputeGame` created with the given parameters.
    ///         Returns `address(0)` if nonexistent.
    function gameAtIndex(uint256 _index)
        external
        view
        returns (GameType gameType_, Timestamp timestamp_, IDisputeGame proxy_);

    /// @notice `gameImpls` is a mapping that maps `GameType`s to their respective
    ///         `IDisputeGame` implementations.
    /// @param _gameType The type of the dispute game.
    /// @return impl_ The address of the implementation of the game type.
    ///         Will be cloned on creation of a new dispute game with the given `gameType`.
    function gameImpls(GameType _gameType) external view returns (IDisputeGame impl_);

    /// @notice Returns the required bonds for initializing a dispute game of the given type.
    /// @param _gameType The type of the dispute game.
    /// @return bond_ The required bond for initializing a dispute game of the given type.
    function initBonds(GameType _gameType) external view returns (uint256 bond_);

    /// @notice Creates a new DisputeGame proxy contract.
    /// @param _gameType The type of the DisputeGame - used to decide the proxy implementation.
    /// @param _rootClaim The root claim of the DisputeGame.
    /// @param _extraData Any extra data that should be provided to the created dispute game.
    /// @return proxy_ The address of the created DisputeGame proxy.
    function create(
        GameType _gameType,
        Claim _rootClaim,
        bytes calldata _extraData
    )
        external
        payable
        returns (IDisputeGame proxy_);

    /// @notice Sets the implementation contract for a specific `GameType`.
    /// @dev May only be called by the `owner`.
    /// @param _gameType The type of the DisputeGame.
    /// @param _impl The implementation contract for the given `GameType`.
    function setImplementation(GameType _gameType, IDisputeGame _impl) external;

    /// @notice Sets the bond (in wei) for initializing a game type.
    /// @dev May only be called by the `owner`.
    /// @param _gameType The type of the DisputeGame.
    /// @param _initBond The bond (in wei) for initializing a game type.
    function setInitBond(GameType _gameType, uint256 _initBond) external;

    /// @notice Returns a unique identifier for the given dispute game parameters.
    /// @dev Hashes the concatenation of `gameType . rootClaim . extraData`
    ///      without expanding memory.
    /// @param _gameType The type of the DisputeGame.
    /// @param _rootClaim The root claim of the DisputeGame.
    /// @param _extraData Any extra data that should be provided to the created dispute game.
    /// @return uuid_ The unique identifier for the given dispute game parameters.
    function getGameUUID(
        GameType _gameType,
        Claim _rootClaim,
        bytes memory _extraData
    )
        external
        pure
        returns (Hash uuid_);

    /// @notice Finds the `_n` most recent `GameId`'s of type `_gameType` starting at `_start`. If there are less than
    ///         `_n` games of type `_gameType` starting at `_start`, then the returned array will be shorter than `_n`.
    /// @param _gameType The type of game to find.
    /// @param _start The index to start the reverse search from.
    /// @param _n The number of games to find.
    function findLatestGames(
        GameType _gameType,
        uint256 _start,
        uint256 _n
    )
        external
        view
        returns (GameSearchResult[] memory games_);
}

DisputeGame Interface

The dispute game interface defines a generic, black-box dispute. It exposes stateful information such as the status of the dispute, when it was created, as well as the bootstrap data and dispute type. This interface exposes one state mutating function, resolve, which when implemented should deterministically yield an opinion about the rootClaim and reflect the opinion by updating the status to CHALLENGER_WINS or DEFENDER_WINS.

Clones of the IDisputeGame's initialize functions will be called by the DisputeGameFactory atomically upon creation.

/// @title IDisputeGame
/// @notice The generic interface for a DisputeGame contract.
interface IDisputeGame is IInitializable {
    /// @notice Emitted when the game is resolved.
    /// @param status The status of the game after resolution.
    event Resolved(GameStatus indexed status);

    /// @notice Returns the timestamp that the DisputeGame contract was created at.
    /// @return createdAt_ The timestamp that the DisputeGame contract was created at.
    function createdAt() external view returns (Timestamp createdAt_);

    /// @notice Returns the timestamp that the DisputeGame contract was resolved at.
    /// @return resolvedAt_ The timestamp that the DisputeGame contract was resolved at.
    function resolvedAt() external view returns (Timestamp resolvedAt_);

    /// @notice Returns the current status of the game.
    /// @return status_ The current status of the game.
    function status() external view returns (GameStatus status_);

    /// @notice Getter for the game type.
    /// @dev The reference impl should be entirely different depending on the type (fault, validity)
    ///      i.e. The game type should indicate the security model.
    /// @return gameType_ The type of proof system being used.
    function gameType() external view returns (GameType gameType_);

    /// @notice Getter for the creator of the dispute game.
    /// @dev `clones-with-immutable-args` argument #1
    /// @return creator_ The creator of the dispute game.
    function gameCreator() external pure returns (address creator_);

    /// @notice Getter for the root claim.
    /// @dev `clones-with-immutable-args` argument #2
    /// @return rootClaim_ The root claim of the DisputeGame.
    function rootClaim() external pure returns (Claim rootClaim_);

    /// @notice Getter for the parent hash of the L1 block when the dispute game was created.
    /// @dev `clones-with-immutable-args` argument #3
    /// @return l1Head_ The parent hash of the L1 block when the dispute game was created.
    function l1Head() external pure returns (Hash l1Head_);

    /// @notice Getter for the extra data.
    /// @dev `clones-with-immutable-args` argument #4
    /// @return extraData_ Any extra data supplied to the dispute game contract by the creator.
    function extraData() external pure returns (bytes memory extraData_);

    /// @notice If all necessary information has been gathered, this function should mark the game
    ///         status as either `CHALLENGER_WINS` or `DEFENDER_WINS` and return the status of
    ///         the resolved game. It is at this stage that the bonds should be awarded to the
    ///         necessary parties.
    /// @dev May only be called if the `status` is `IN_PROGRESS`.
    /// @return status_ The status of the game after resolution.
    function resolve() external returns (GameStatus status_);

    /// @notice A compliant implementation of this interface should return the components of the
    ///         game UUID's preimage provided in the cwia payload. The preimage of the UUID is
    ///         constructed as `keccak256(gameType . rootClaim . extraData)` where `.` denotes
    ///         concatenation.
    /// @return gameType_ The type of proof system being used.
    /// @return rootClaim_ The root claim of the DisputeGame.
    /// @return extraData_ Any extra data supplied to the dispute game contract by the creator.
    function gameData() external view returns (GameType gameType_, Claim rootClaim_, bytes memory extraData_);
}

Fault Dispute Game

Table of Contents

• Overview
• Definitions
  • Virtual Machine (VM)
  • PreimageOracle
  • Execution Trace
  • Claims
  • Anchor State
  • Anchor State Registry
  • DAG
  • Subgame
  • Game Tree
  • Position
  • MAX_CLOCK_DURATION
  • CLOCK_EXTENSION
  • Freeloader Claims
• Core Game Mechanics
  • Actors
  • Moves
    • Attack
    • Defend
  • L2 Block Number Challenge
  • Step
  • Step Types
  • PreimageOracle Interaction
  • Team Dynamics
  • Game Clock
  • Resolution
    • Resolving the L2 Block Number Challenge
  • Finalization

Overview

The Fault Dispute Game (FDG) is a specific type of dispute game that verifies the validity of a root claim by iteratively bisecting over output roots and execution traces of single block state transitions down to a single instruction step. It relies on a Virtual Machine (VM) to falsify invalid claims made at a single instruction step.

Actors, i.e. Players, interact with the game by making claims that dispute other claims in the FDG. Each claim made narrows the range over the entire historical state of L2, until the source of dispute is a single state transition. Once a time limit is reached, the dispute game is resolved, based on claims made that are disputed and which aren't, to determine the winners of the game.

Definitions

Virtual Machine (VM)

This is a state transition function (STF) that takes a pre-state and computes the post-state. The VM may access data referenced during the STF and as such, it also accepts a proof of this data. Typically, the pre-state contains a commitment to the proof to verify the integrity of the data referenced.

Mathematically, we define the STF as $VM(S_i,P_i)$ where

• $S_i$ is the pre-state
• $P_i$ is an optional proof needed for the transition from $S_i$ to $S_{i+1}$.

PreimageOracle

This is a pre-image data store. It is often used by VMs to read external data during its STF. Before successfully executing a VM STF, it may be necessary to preload the PreimageOracle with pertinent data. The method for key-based retrieval of these pre-images varies according to the specific VM.

Execution Trace

An execution trace $T$ is a sequence $(S_0,S_1,S_2,...,S_n)$ where each $S_i$ is a VM state and for each $i$, $0\le i<n$, $S_{i+1}=VM(S_i,P_i)$. Every execution trace has a unique starting state, $S_0$, that's preset to a FDG implementation. We refer to this state as the ABSOLUTE_PRESTATE.

Claims

Claims assert an output root or the state of the FPVM at a given instruction. This is represented as a Hash type, a bytes32 representing either an output root or a commitment to the last VM state in a trace. A FDG is initialized with an output root that corresponds to the state of L2 at a given L2 block number, and execution trace subgames at SPLIT_DEPTH + 1 are initialized with a claim that commits to the entire execution trace between two consecutive output roots (a block n -> n+1 state transition). As we'll see later, there can be multiple claims, committing to different output roots and FPVM states in the FDG.

Anchor State

An anchor state, or anchor output root, is a previous output root that is assumed to be valid. An FDG is always initialized with an anchor state and execution is carried out between this anchor state and the claimed output root. FDG contracts pull their anchor state from the Anchor State Registry contract. The initial anchor state for a FDG is the genesis state of the L2.

Clients must currently gather L1 data for the window between the anchor state and the claimed state. In order to reduce this L1 data requirement, claims about the state of the L2 become new anchor states when dispute games resolve in their favor. FDG contracts set their anchor states at initialization time so that these updates do not impact active games.

Anchor State Registry

The Anchor State Registry is a registry that maps FDG types to their current anchor states. The Anchor State Registry is specific to Fault Dispute Game contracts and may not be applicable to other types of dispute game contracts that do not have the same concept of state that progresses over time.

DAG

A Directed Acyclic Graph $G=(V,E)$ representing the relationship between claims, where:

• $V$ is the set of nodes, each representing a claim. Formally, $V=\{C_1,C_2,...,C_n\}$, where $C_i$ is a claim.
• $E$ is the set of directed edges. An edge $(C_i,C_j)$ exists if $C_j$ is a direct dispute against $C_i$ through either an "Attack" or "Defend" move.

Subgame

A sub-game is a DAG of depth 1, where the root of the DAG is a Claim and the children are Claims that counter the root. A good mental model around this structure is that it is a fundamental dispute between two parties over a single piece of information. These subgames are chained together such that a child within a subgame is the root of its own subgame, which is visualized in the resolution section. There are two types of sub-games in the fault dispute game:

1. Output Roots
2. Execution Trace Commitments

At and above the split depth, all subgame roots correspond to output roots, or commitments to the full state of L2 at a given L2 block number. Below the split depth, subgame roots correspond to commitments to the fault proof VM's state at a given instruction step.

Game Tree

The Game Tree is a binary tree of positions. Every claim in the DAG references a position in the Game Tree. The Game Tree has a split depth and maximum depth, SPLIT_DEPTH and MAX_GAME_DEPTH respectively, that are both preset to an FDG implementation. The split depth defines the maximum depth at which claims about output roots can occur, and below it, execution trace bisection occurs. Thus, the Game Tree contains $2^{d+1}-1$ positions, where $d$ is the MAX_GAME_DEPTH (unless $d=0$, in which case there's only 1 position).

The full game tree, with a layer of the tree allocated to output bisection, and sub-trees after an arbitrary split depth, looks like:

Position

A position represents the location of a claim in the Game Tree. This is represented by a "generalized index" (or gindex) where the high-order bit is the level in the tree and the remaining bits is a unique bit pattern, allowing a unique identifier for each node in the tree.

The gindex of a position $n$ can be calculated as $2^{d(n)}+idx(n)$, where:

• $d(n)$ is a function returning the depth of the position in the Game Tree
• $idx(n)$ is a function returning the index of the position at its depth (starting from the left).

Positions at the deepest level of the game tree correspond to indices in the execution trace, whereas claims at the split depth represent single L2 blocks' output roots. Positions higher up the game tree also cover the deepest, right-most positions relative to the current position. We refer to this coverage as the trace index of a Position.

This means claims commit to an execution trace that terminates at the same index as their Position's trace index. That is, for a given trace index $n$, its state witness hash corresponds to the $S_n$ th state in the trace.

Note that there can be multiple positions covering the same trace index.

MAX_CLOCK_DURATION

This is an immutable, preset to a FDG implementation, representing the maximum amount of time that may accumulate on a team's chess clock.

CLOCK_EXTENSION

This is an immutable, preset to a FDG implementation, representing the flat credit that is given to a team's clock if their clock has less than CLOCK_EXTENSION seconds remaining.

Freeloader Claims

Due to the subgame resolution logic, there are certain moves which result in the correct final resolution of the game, but do not pay out bonds to the correct parties.

An example of this is as follows:

1. Alice creates a dispute game with an honest root claim.
2. Bob counters the honest root with a correct claim at the implied L2 block number.
3. Alice performs a defense move against Bob's counter, as the divergence exists later in Bob's view of the chain state.
4. Bob attacks his own claim.

Bob's attack against his own claim is a counter to a bad claim, but with the incorrect pivot direction. If left untouched, because it exists at a position further left than Alice's, he will reclaim his own bond upon resolution. Because of this, the honest challenger must always counter freeloader claims for incentive compatibility to be preserved.

Critically, freeloader claims, if left untouched, do not influence incorrect resolution of the game globally.

Core Game Mechanics

This section specifies the core game mechanics of the FDG. The full FDG mechanics includes a specification for Bonds. Readers should understand basic game mechanics before reading up on the Bond specification.

Actors

The game involves two types of participants (or Players): Challengers and Defenders. These players are grouped into separate teams, each employing distinct strategies to interact with the game. Team members share a common goal regarding the game's outcome. Players interact with the game primarily through moves.

Moves

A Move is a challenge against an existing claim and must include an alternate claim asserting a different trace. Moves can either be attacks or defenses and serve to update to DAG by adding nodes and edges targeting the disputed claim.

Moves within the fault dispute game can claim two separate values: output roots and execution trace commitments. At and above the SPLIT_DEPTH, claims correspond to output roots, while below the split depth, they correspond to execution trace commitments.

Initially, claims added to the DAG are uncontested (i.e. not countered). Once a move targets a claim, that claim is considered countered. The status of a claim — whether it's countered or not — helps determine its validity and, ultimately, the game's winner.

Attack

A logical move made when a claim is disagreed with. A claim at the relative attack position to a node, n, in the Game Tree commits to half of the trace of the n’s claim. The attack position relative to a node can be calculated by multiplying its gindex by 2.

To illustrate this, here's a Game Tree highlighting an attack on a Claim positioned at 6.

Attacking the node at 6 moves creates a new claim positioned at 12.

Defend

The logical move against a claim when you agree with both it and its parent. A defense at the relative position to a node, n, in the Game Tree commits to the first half of n + 1’s trace range.

Note that because of this, some nodes may never exist within the Game Tree. However, they're not necessary as these nodes have complimentary, valid positions with the same trace index within the tree. For example, a Position with gindex 5 has the same trace index as another Position with gindex 2. We can verify that all trace indices have valid moves within the game:

There may be multiple claims at the same position, so long as their state witness hashes are unique.

Each move adds new claims to the Game Tree at strictly increasing depth. Once a claim is at MAX_GAME_DEPTH, the only way to dispute such claims is to step.

L2 Block Number Challenge

This is a special type of action, made by the Challenger, to counter a root claim.

Given an output root preimage and its corresponding RLP-encoded L2 block header, the L2 block number can be verified. This process ensures the integrity and authenticity of an L2 block number. The procedure for this verification involves three steps: checking the output root preimage, validating the block hash preimage, and extracting the block number from the RLP-encoded header. By comparing the challenger-supplied preimages and the extracted block number against their claimed values, the consistency of the L2 block number with the one in the provided header can be confirmed, detecting any discrepancies.

Root claims made with an invalid L2 block number can be disputed through a special challenge. This challenge is validated in the FDG contract using the aforementioned procedure. However, it is crucial to note that this challenge can only be issued against the root claim, as it's the only entity making explicit claims on the L2 block number. A successful challenge effectively disputes the root claim once its subgame is resolved.

Step

At MAX_GAME_DEPTH, the position of claims correspond to indices of an execution trace. It's at this point that the FDG is able to query the VM to determine the validity of claims, by checking the states they're committing to. This is done by applying the VM's STF to the state a claim commits to. If the STF post-state does not match the claimed state, the challenge succeeds.

/// @notice Perform an instruction step via an on-chain fault proof processor.
/// @dev This function should point to a fault proof processor in order to execute
///      a step in the fault proof program on-chain. The interface of the fault proof
///      processor contract should adhere to the `IBigStepper` interface.
/// @param _claimIndex The index of the challenged claim within `claimData`.
/// @param _isAttack Whether or not the step is an attack or a defense.
/// @param _stateData The stateData of the step is the preimage of the claim at the given
///        prestate, which is at `_stateIndex` if the move is an attack and `_claimIndex` if
///        the move is a defense. If the step is an attack on the first instruction, it is
///        the absolute prestate of the fault proof VM.
/// @param _proof Proof to access memory nodes in the VM's merkle state tree.
function step(uint256 _claimIndex, bool _isAttack, bytes calldata _stateData, bytes calldata _proof) external;

Step Types

Similar to moves, there are two ways to step on a claim; attack or defend. These determine the pre-state input to the VM STF and the expected output.

• Attack Step - Challenges a claim by providing a pre-state, proving an invalid state transition. It uses the previous state in the execution trace as input and expects the disputed claim's state as output. There must exist a claim in the DAG that commits to the input.
• Defense Step - Challenges a claim by proving it was an invalid attack, thereby defending the disputed ancestor's claim. It uses the disputed claim's state as input and expects the next state in the execution trace as output. There must exist a claim in the DAG that commits to the expected output.

The FDG step handles the inputs to the VM and asserts the expected output. A step that successfully proves an invalid post-state (when attacking) or pre-state (when defending) is a successful counter against the disputed claim. Players interface with step by providing an indicator of attack and state data (including any proofs) that corresponds to the expected pre/post state (depending on whether it's an attack or defend). The FDG will assert that an existing claim commits to the state data provided by players.

PreimageOracle Interaction

Certain steps (VM state transitions) require external data to be available by the PreimageOracle. To ensure a successful state transition, players should provide this data in advance. The FDG provides the following interface to manage data loaded to the PreimageOracle:

/// @notice Posts the requested local data to the VM's `PreimageOracle`.
/// @param _ident The local identifier of the data to post.
/// @param _execLeafIdx The index of the leaf claim in an execution subgame that requires the local data for a step.
/// @param _partOffset The offset of the data to post.
function addLocalData(uint256 _ident, uint256 _execLeafIdx, uint256 _partOffset) external;

The addLocalData function loads local data into the VM's PreimageOracle. This data consists of bootstrap data for the program. There are multiple sets of local preimage keys that belong to the FaultDisputeGame contract due to the ability for players to bisect to any block $n\to n+1$ state transition since the configured anchor state, the _execLeafIdx parameter enables a search for the starting / disputed outputs to be performed such that the contract can write to and reference unique local keys in the PreimageOracle for each of these $n\to n+1$ transitions.

For global keccak256 preimages, there are two routes for players to submit:

1. Small preimages atomically.
2. Large preimages via streaming.

Global keccak256 preimages are non-context specific and can be submitted directly to the PreimageOracle via the loadKeccak256PreimagePart function, which takes the part offset as well as the full preimage. In the event that the preimage is too large to be submitted through calldata in a single block, challengers must resort to the streaming option.

Large Preimage Proposals

Large preimage proposals allow for submitters to stream in a large preimage over multiple transactions, along-side commitments to the intermediate state of the keccak256 function after absorbing/permuting the $1088$ bit block. This data is progressively merkleized on-chain as it is streamed in, with each leaf constructed as follows:

/// @notice Returns a leaf hash to add to a preimage proposal merkle tree.
/// @param input A single 136 byte chunk of the input.
/// @param blockIndex The index of the block that `input` corresponds to in the full preimage's absorption.
/// @param stateCommitment The hash of the full 5x5 state matrix *after* absorbing and permuting `input`.
function hashLeaf(
    bytes memory input,
    uint256 blockIndex,
    bytes32 stateCommitment
) internal view returns (bytes32 leaf) {
    require(input.length == 136, "input must be exactly the size of the keccak256 rate");

    leaf = keccak256(abi.encodePacked(input, blockIndex, stateCommitment));
}

Once the full preimage and all intermediate state commitments have been posted, the large preimage proposal enters a challenge period. During this time, a challenger can reconstruct the merkle tree that was progressively built on-chain locally by scanning the block bodies that contain the proposer's leaf preimages. If they detect that a commitment to the intermediate state of the hash function is incorrect at any step, they may perform a single-step dispute for the proposal in the PreimageOracle. This involves:

1. Creating a merkle proof for the agreed upon prestate leaf (not necessary if the invalid leaf is the first one, the setup state of the matrix is constant.) within the proposal's merkle root.
2. Creating a merkle proof for the disputed post state leaf within the proposal's merkle root.
3. Computing the state matrix at the agreed upon prestate (not necessary if the invalid leaf is the first one, the setup state of the matrix is constant.)

The challenger then submits this data to the PreimageOracle, where the post state leaf's claimed input is absorbed into the pre state leaf's state matrix and the SHA3 permutation is executed on-chain. After that, the resulting state matrix is hashed and compared with the proposer's claim in the post state leaf. If the hash does not match, the proposal is marked as challenged, and it may not be finalized. If, after the challenge period is concluded, a proposal has no challenges, it may be finalized and the preimage part may be placed into the authorized mappings for the FPVM to read.

Team Dynamics

Challengers seek to dispute the root claim, while Defenders aim to support it. Both types of actors will move accordingly to support their team. For Challengers, this means attacking the root claim and disputing claims positioned at even depths in the Game Tree. Defenders do the opposite by disputing claims positioned at odd depths.

Players on either team are motivated to support the actions of their teammates. This involves countering disputes against claims made by their team (assuming these claims are honest). Uncontested claims are likely to result in a loss, as explained later under Resolution.

Game Clock

Every claim in the game has a Clock. A claim inherits the clock of its grandparent claim in the DAG (and so on). Akin to a chess clock, it keeps track of the total time each team takes to make moves, preventing delays. Making a move resumes the clock for the disputed claim and pauses it for the newly added one.

If a move is performed, where the potential grandchild's clock has less time than CLOCK_EXTENSION seconds remaining, the potential grandchild's clock is granted exactly CLOCK_EXTENSION seconds remaining. This is to combat the situation where a challenger must inherit a malicious party's clock when countering a freeloader claim, in order to preserve incentive compatibility for the honest party. As the extension only applies to the potential grandchild's clock, the max possible extension for the game is bounded, and scales with the MAX_GAME_DEPTH.

If the potential grandchild is an execution trace bisection root claim and their clock has less than CLOCK_EXTENSION seconds remaining, exactly CLOCK_EXTENSION * 2 seconds are allocated for the potential grandchild. This extra time is alloted to allow for completion of the off-chain FPVM run to generate the initial instruction trace.

A move against a particular claim is no longer possible once the parent of the disputed claim's Clock has accumulated MAX_CLOCK_DURATION seconds. By which point, the claim's clock has expired.

Resolution

Resolving the FDG determines which team won the game. To do this, we use the internal sub game structure. Each claim within the game is the root of its own sub game. These subgames are modeled as nested DAGs, each with a max depth of 1. In order for a claim to be considered countered, only one of its children must be uncountered. Subgames can also not be resolved until all of their children, which are subgames themselves, have been resolved and the potential opponent's chess clock has run out. To determine if the potential opponent's chess clock has ran out, and therefore no more moves against the subgame are possible, the duration elapsed on the subgame root's parent clock is added to the difference between the current time and the timestamp of the subgame root's creation. Because each claim is the root of its own sub-game, truth percolates upwards towards the root claim by resolving each individual sub-game bottom-up.

In a game like the one below, we can resolve up from the deepest subgames. Here, we'd resolve b0 to uncountered and a0 to countered by walking up from their deepest children, and once all children of the root game are recursively resolved, we can resolve the root to countered due to b0 remaining uncountered.

https://github.com/ethereum-optimism/optimism/assets/8406232/d2b708a0-539e-439d-96bd-c2f66f3a45f8

Another example is this game, which has a slightly different structure. Here, the root claim will also be countered due to b0 remaining uncountered.

Given these rules, players are motivated to move quickly to challenge all dishonest claims. Each move bisects the historical state of L2 and eventually, MAX_GAME_DEPTH is reached where disputes can be settled conclusively. Dishonest players are disincentivized to participate, via backwards induction, as an invalid claim won't remain uncontested. Further incentives can be added to the game by requiring claims to be bonded, while rewarding game winners using the bonds of dishonest claims.

Resolving the L2 Block Number Challenge

The resolution of an L2 block number challenge occurs in the same manner as subgame resolution, with one caveat; the L2 block number challenger, if it exist, must be the winner of a root subgame. Thus, no moves against the root, including uncontested ones, can win a root subgame that has an L2 block number challenge.

Finalization

Once the game is resolved, if the claim is shown to be valid, the FDG reports its state to the Anchor State Registry. The Anchor State Registry verifies that the request to update an anchor state comes from a FDG contract created by the Dispute Game Factory contract, confirms that the game resolved in favor of the defender, and confirms that the updated state would be newer than the current anchor state (based on the result of FaultDisputeGame.l2BlockNumber()). If these conditions are true, the Anchor State Registry updates the anchor state for the given game type.

Note that the dependency of the Anchor State Registry on the l2BlockNumber() function means that the registry may not be applicable to dispute game types that do not have a similar sense of progressing state over time. Currently, the Anchor State Registry is therefore assumed to be specific to the Fault Dispute Game contract and not to other dispute game types.

Honest Challenger (Fault Dispute Game)

Table of Contents

• Overview
• Invariants
• Fault Dispute Game Responses
  • Moves
  • Steps
  • Timeliness
• Resolution

Overview

The honest challenger is an agent interacting in the Fault Dispute Game that supports honest claims and disputes false claims. An honest challenger strives to ensure a correct, truthful, game resolution. The honest challenger is also rational as any deviation from its behavior will result in negative outcomes. This document specifies the expected behavior of an honest challenger.

The Honest Challenger has two primary duties:

1. Support valid root claims in Fault Dispute Games.
2. Dispute invalid root claims in Fault Dispute Games.

The honest challenger polls the DisputeGameFactory contract for new and on-going Fault Dispute Games. For verifying the legitimacy of claims, it relies on a synced, trusted rollup node as well as a trace provider (ex: Cannon). The trace provider must be configured with the ABSOLUTE_PRESTATE of the game being interacted with to generate the traces needed to make truthful claims.

Invariants

To ensure an accurate and incentive compatible fault dispute system, the honest challenger behavior must preserve three invariants for any game:

1. The game resolves as DefenderWins if the root claim is correct and ChallengerWins if the root claim is incorrect
2. The honest challenger is refunded the bond for every claim it posts and paid the bond of the parent of that claim
3. The honest challenger never counters its own claim

Fault Dispute Game Responses

The honest challenger determines which claims to counter by iterating through the claims in the order they are stored in the contract. This ordering ensures that a claim's ancestors are processed prior to the claim itself. For each claim, the honest challenger determines and tracks the set of honest responses to all claims, regardless of whether that response already exists in the full game state.

The root claim is considered to be an honest claim if and only if it has a state witness Hash that agrees with the honest challenger's state witness hash for the root claim.

The honest challenger should counter a claim if and only if:

1. The claim is a child of a claim in the set of honest responses
2. The set of honest responses, contains a sibling to the claim with a trace index greater than or equal to the claim's trace index

Note that this implies the honest challenger never counters its own claim, since there is at most one honest counter to each claim, so an honest claim never has an honest sibling.

Moves

To respond to a claim with a depth in the range of [1, MAX_DEPTH], the honest challenger determines if the claim has a valid commitment. If the state witness hash matches the honest challenger's at the same trace index, then we disagree with the claim's stance by move to defend. Otherwise, the claim is attacked.

The claim that would be added as a result of the move is added to the set of honest moves being tracked.

If the resulting claim does not already exist in the full game state, the challenger issue the move by calling the FaultDisputeGame contract.

Steps

At the max depth of the game, claims represent commitments to the state of the fault proof VM at a single instruction step interval. Because the game can no longer bisect further, when the honest challenger counters these claims, the only option for an honest challenger is to execute a VM step on-chain to disprove the claim at MAX_GAME_DEPTH.

If the counteredBy of the claim being countered is non-zero, the claim has already been countered and the honest challenger does not perform any action.

Otherwise, similar to the above section, the honest challenger will issue an attack step when in response to such claims with invalid state witness commitments. Otherwise, it issues a defense step.

Timeliness

The honest challenger responds to claims as soon as possible to avoid the clock of its counter-claim from expiring.

Resolution

When the chess clock of a subgame root has run out, the subgame can be resolved. The honest challenger should resolve all subgames in bottom-up order, until the subgame rooted at the game root is resolved.

The honest challenger accomplishes this by calling the resolveClaim function on the FaultDisputeGame contract. Once the root claim's subgame is resolved, the challenger then finally calls the resolve function to resolve the entire game.

The FaultDisputeGame does not put a time cap on resolution - because of the liveness assumption on honest challengers and the bonds attached to the claims they’ve countered, challengers are economically incentivized to resolve the game promptly to capture the bonds.

Bond Incentives

Table of Contents

• Overview
• Moves
• Subgame Resolution
  • Leftmost Claim Incentives
• Fault Proof Mainnet Incentives
  • Authenticated Roles
  • Base Fee Assumption
  • Bond Scaling
  • Required Bond Formula
  • Other Incentives
  • DelayedWETH
    • Sub-Account Model
    • Delay Period
    • Integration

Overview

Bonds is an add-on to the core Fault Dispute Game. The core game mechanics are designed to ensure honesty as the best response to winning subgames. By introducing financial incentives, Bonds makes it worthwhile for honest challengers to participate. Without the bond reward incentive, the FDG will be too costly for honest players to participate in given the cost of verifying and making claims.

Implementations may allow the FDG to directly receive bonds, or delegate this responsibility to another entity. Regardless, there must be a way for the FDG to query and distribute bonds linked to a claim.

Bonds are integrated into the FDG in two areas:

• Moves
• Subgame Resolution

Moves

Moves must be adequately bonded to be added to the FDG. This document does not specify a scheme for determining the minimum bond requirement. FDG implementations should define a function computing the minimum bond requirement with the following signature:

function getRequiredBond(Position _movePosition) public pure returns (uint256 requiredBond_)

As such, attacking or defending requires a check for the getRequiredBond() amount against the bond attached to the move. To incentivize participation, the minimum bond should cover the cost of a possible counter to the move being added. Thus, the minimum bond depends only on the position of the move that's added.

Subgame Resolution

If a subgame root resolves incorrectly, then its bond is distributed to the leftmost claimant that countered it. This creates an incentive to identify the earliest point of disagreement in an execution trace. The subgame root claimant gets back its bond iff it resolves correctly.

At maximum game depths, where a claimant counters a bonded claim via step, the bond is instead distributed to the account that successfully called step.

Leftmost Claim Incentives

There exists defensive positions that cannot be countered, even if they hold invalid claims. These positions are located on the same level as honest claims, but situated to its right (i.e. its gindex > honest claim's).

An honest challenger can always successfully dispute any sibling claims not positioned to the right of an honest claim. The leftmost payoff rule encourages such disputes, ensuring only one claim is leftmost at correct depths. This claim will be the honest one, and thus bond rewards will be directed exclusively to honest claims.

Fault Proof Mainnet Incentives

This section describes the specific bond incentives to be used for the Fault Proof Mainnet launch of the OP Stack fault proof system.

Authenticated Roles

Base Fee Assumption

FPM bonds are to assume a fixed 200 Gwei base fee. Future iterations of the fault proof may include a dynamic base fee calculation. For the moment, we suppose that the Guardian address may account for increased average base fees by updating the OptimismPortal contract to a new respected game type with a higher assumed base fee.

Bond Scaling

FPM bonds are priced in the amount of gas that they are intended to cover. Bonds start at the very first depth of the game at a baseline of 400_000 gas. The 400_000 value is chosen as a deterrence amount that is approximately double the cost to respond at the top level. Bonds scale up to a value of 300_000_000 gas, a value chosen to cover approximately double the cost of a max-size Large Preimage Proposal.

We use a multiplicative scaling mechanism to guarantee that the ratio between bonds remains constant. We determine the multiplier based on the proposed MAX_DEPTH of 73. We can use the formula x = (300_000_000 / 400_000) ** (1 / 73) to determine that x = 1.09493. At each depth N, the amount of gas charged is therefore 400_000 * (1.09493 ** N)

Below is a diagram demonstrating this curve for a max depth of 73.

Required Bond Formula

Applying the Base Fee Assumption and Bond Scaling specifications, we have a getRequiredBond function:

def get_required_bond(position):
    assumed_gas_price = 200 gwei
    base_gas_charged = 400_000
    gas_charged = 400_000 * (1.09493 ** position.depth)
    return gas_charged * assumed_gas_price

Other Incentives

There are other costs associated with participating in the game, including operating a challenger agent and the opportunity cost of locking up capital in the dispute game. While we do not explicitly create incentives to cover these costs, we assume that the current bond rewards, based on this specification, are enough as a whole to cover all other costs of participation.

DelayedWETH

FPM introduces a contract DelayedWETH designed to hold the bonded ETH for each Fault Dispute Game. DelayedWETH is an extended version of the standard WETH contract that introduces a delayed unwrap mechanism that allows an owner address to function as a backstop in the case that a Fault Dispute Game would incorrectly distribute bonds.

DelayedWETH is modified from WETH as follows:

• DelayedWETH is an upgradeable proxy contract.
• DelayedWETH has an owner() address. We typically expect this to be set to the System Owner address.
• DelayedWETH has a delay() function that returns a period of time that withdrawals will be delayed.
• DelayedWETH has an unlock(guy,wad) function that modifies a mapping called withdrawals keyed as withdrawals[msg.sender][guy] => WithdrawalRequest where WithdrawalRequest is struct Withdrawal Request { uint256 amount, uint256 timestamp }. When unlock is called, the timestamp for withdrawals[msg.sender][guy] is set to the current timestamp and the amount is increased by the given amount.
• DelayedWETH modifies the WETH.withdraw function such that an address must provide a "sub-account" to withdraw from. The function signature becomes withdraw(guy,wad). The function retrieves withdrawals[msg.sender][guy] and checks that the current block.timestamp is greater than the timestamp on the withdrawal request plus the delay() seconds and reverts if not. It also confirms that the amount being withdrawn is less than the amount in the withdrawal request. Before completing the withdrawal, it reduces the amount contained within the withdrawal request. The original withdraw(wad) function becomes an alias for withdraw(msg.sender, wad). withdraw(guy,wad) will not be callable when SuperchainConfig.paused() is true.
• DelayedWETH has a hold() function that allows the owner() address to give itself an allowance from any address.
• DelayedWETH has a recover() function that allows the owner() address to recover any amount of ETH from the contract.

Sub-Account Model

This specification requires that withdrawal requests specify "sub-accounts" that these requests correspond to. This takes the form of requiring that unlock and withdraw both take an address guy parameter as input. By requiring this extra input, withdrawals are separated between accounts and it is always possible to see how much WETH a specific end-user of the FaultDisputeGame can withdraw at any given time. It is therefore possible for the DelayedWETH contract to account for all bug cases within the FaultDisputeGame as long as the FaultDisputeGame always passes the correct address into withdraw.

Delay Period

We propose a delay period of 7 days for most OP Stack chains. 7 days provides sufficient time for the owner() of the DelayedWETH contract to act even if that owner is a large multisig that requires action from many different members over multiple timezones.

Integration

DelayedWETH is expected to be integrated into the Fault Dispute Game as follows:

• When FaultDisputeGame.initialize is triggered, DelayedWETH.deposit{value: msg.value}() is called.
• When FaultDisputeGame.move is triggered, DelayedWETH.deposit{value: msg.value}() is called.
• When FaultDisputeGame.resolveClaim is triggered, DelayedWETH.unlock(recipient, bond) is called.
• When FaultDisputeGame.claimCredit is triggered, DelayedWETH.withdraw(recipient, claim) is called.

sequenceDiagram
    participant U as User
    participant FDG as FaultDisputeGame
    participant DW as DelayedWETH

    U->>FDG: initialize()
    FDG->>DW: deposit{value: msg.value}()
    Note over DW: FDG gains balance in DW

    loop move by Users
        U->>FDG: move()
        FDG->>DW: deposit{value: msg.value}()
        Note over DW: Increases FDG balance in DW
    end

    loop resolveClaim by Users
        U->>FDG: resolveClaim()
        FDG->>DW: unlock(recipient, bond)
        Note over DW: Starts timer for recipient
    end

    loop claimCredit by Users
        U->>FDG: claimCredit()
        FDG->>DW: withdraw(recipient, claim)
        Note over DW: Checks timer/amount for recipient
        DW->>FDG: Transfer claim to FDG
        FDG->>U: Transfer claim to User
    end

Bridge Integration

Table of Contents

• Overview
• Legacy Semantics
• FPAC OptimismPortal Mods Specification
  • Roles - OptimismPortal
  • New DeployConfig Variables
  • Data Structures
  • State Layout
    • Legacy Spacers
    • New State Variables
  • proveWithdrawalTransaction modifications
    • Interface
    • New Invariants - proveWithdrawalTransaction
    • Changed Invariants - proveWithdrawalTransaction
  • finalizeWithdrawalTransaction modifications
    • New Invariants - finalizeWithdrawalTransaction
    • Changed Invariants - finalizeWithdrawalTransaction
  • Air-gap
    • Blacklisting DisputeGames
    • Blacklisting a full GameType
  • Proxy Upgrade
• Permissioned FaultDisputeGame
  • Roles - PermissionedDisputeGame
  • Modifications

Overview

With fault proofs, the withdrawal path changes such that withdrawals submitted to the OptimismPortal are proven against output proposals submitted as a FaultDisputeGame prior to being finalized. Output proposals are now finalized whenever a dispute game resolves in their favor.

Legacy Semantics

The OptimismPortal uses the L2OutputOracle in the withdrawal path of the rollup to allow users to prove the presence of their withdrawal inside of the L2ToL1MessagePasser account storage root, which can be retrieved by providing a preimage to an output root in the oracle. The oracle currently holds a list of all L2 outputs proposed to L1 by a permissioned PROPOSER key. The list in the contract has the following properties:

• It must always be sorted by the L2 Block Number that the output proposal is claiming it corresponds to.
• All outputs in the list that are > FINALIZATION_PERIOD_SECONDS old are considered "finalized." The separator between unfinalized/finalized outputs moves forwards implicitly as time passes.

Currently, if there is a faulty output proposed by the permissioned PROPOSER key, a separate permissioned CHALLENGER key may intervene. Note that the CHALLENGER role effectively has god-mode privileges, and can currently act without proving that the outputs they're deleting are indeed incorrect. By deleting an output proposal, the challenger also deletes all output proposals in front of it.

With the upgrade to the Fault Proof Alpha Chad system, output proposals are no longer sent to the L2OutputOracle, but to the DisputeGameFactory in order to be fault proven. In contrast to the L2OO, an incorrect output proposal is not deleted, but proven to be incorrect. The semantics of finalization timelines and the definition of a "finalized" output proposal also change. Since the DisputeGameFactory fulfills the same role as the L2OutputOracle in a post fault proofs world by tracking proposed outputs, and the L2OO's semantics are incompatible with the new system, the L2OO is no longer required.

FPAC OptimismPortal Mods Specification

Roles - OptimismPortal

• Guardian: Permissioned actor able to pause the portal, blacklist dispute games, and change the RESPECTED_GAME_TYPE.

New DeployConfig Variables

Data Structures

Withdrawals are now proven against dispute games, which have immutable "root claims" representing the output root being proposed. The ProvenWithdrawal struct is now defined as:

/// @notice Represents a proven withdrawal.
/// @custom:field disputeGameProxy The address of the dispute game proxy that the withdrawal was proven against.
/// @custom:field timestamp        Timestamp at which the withdrawal was proven.
struct ProvenWithdrawal {
    IDisputeGame disputeGameProxy;
    uint64 timestamp;
}

State Layout

Legacy Spacers

Spacers should be added at the following storage slots in the OptimismPortal so that they may not be reused:

New State Variables

DisputeGameFactory address

/// @notice Address of the DisputeGameFactory.
/// @custom:network-specific
DisputeGameFactory public disputeGameFactory;

Respected Game Type

/// @notice The respected game type of the `OptimismPortal`.
///         Can be changed by Guardian.
GameType public respectedGameType;

Respected Game Type Updated Timestamp

/// @notice The timestamp at which the respected game type was last updated.
uint64 public respectedGameTypeUpdatedAt;

New ProvenWithdrawals mapping

/// @notice A mapping of withdrawal hashes to `ProvenWithdrawal` data.
mapping(bytes32 => ProvenWithdrawal) public provenWithdrawals;

Blacklisted DisputeGame mapping

/// @notice A mapping of dispute game addresses to whether or not they are blacklisted.
mapping(IDisputeGame => bool) public disputeGameBlacklist;

proveWithdrawalTransaction modifications

Proving a withdrawal transaction now proves against an output root in a dispute game, rather than one in the L2OutputOracle.

Interface

The type signature of the function does not change, but the purpose of the second argument transitions from providing an index within the L2OutputOracle's l2Outputs array to an index within the DisputeGameFactory's list of created games.

/// @notice Proves a withdrawal transaction.
/// @param _tx               Withdrawal transaction to finalize.
/// @param _disputeGameIndex Index of the dispute game to prove the withdrawal against.
/// @param _outputRootProof  Inclusion proof of the L2ToL1MessagePasser contract's storage root.
/// @param _withdrawalProof  Inclusion proof of the withdrawal in L2ToL1MessagePasser contract.
function proveWithdrawalTransaction(
    Types.WithdrawalTransaction memory _tx,
    uint256 _disputeGameIndex,
    Types.OutputRootProof calldata _outputRootProof,
    bytes[] calldata _withdrawalProof
) external whenNotPaused;

New Invariants - proveWithdrawalTransaction

Trusted GameType The DisputeGameFactory can create many different types of dispute games, delineated by their GameType. The game type of the dispute game fetched from the factory's list at _disputeGameIndex must be of type RESPECTED_GAME_TYPE. The call should revert on all other game types it encounters.

Changed Invariants - proveWithdrawalTransaction

Re-proving withdrawals Users being able to re-prove withdrawals, in special cases, is still necessary to prevent user withdrawals from being bricked. It is kept to protect honest users when they prove their withdrawal inside of a malicious proposal. The timestamp of re-proven withdrawals is still reset.

1. Old: Re-proving is allowed if the output root at the proven withdrawal's l2OutputIndex changed in the L2OutputOracle.
2. New: Re-proving is allowed at any time by the user. When a withdrawal is re-proven, its proof maturity delay is reset.

finalizeWithdrawalTransaction modifications

Finalizing a withdrawal transaction now references a DisputeGame to determine the status of the output proposal that the withdrawal was proven against.

New Invariants - finalizeWithdrawalTransaction

Trusted GameType The DisputeGameFactory can create many different types of dispute games, delineated by their GameType. The game type of the dispute game fetched from the factory's list at _disputeGameIndex must be of type RESPECTED_GAME_TYPE. The call should revert on all other game types it encounters.

Respected Game Type Updated A withdrawal may never be finalized if the dispute game was created before the respected game type was last updated.

Dispute Game Blacklist The Guardian role can blacklist certain DisputeGame addresses in the event of a system failure. If the address of the dispute game that the withdrawal was proven against is present in the disputeGameBlacklist mapping, the call should always revert.

Dispute Game Maturity See "Air-gap"

Changed Invariants - finalizeWithdrawalTransaction

Output Proposal Validity Instead of checking if the proven withdrawal's output proposal has existed for longer the legacy finalization period, we check if the dispute game has resolved in the root claim's favor. A FaultDisputeGame must never be considered to have resolved in the rootClaim's favor unless its status() is equal to DEFENDER_WINS.

Air-gap

Given it's own section due to it's importance, the air gap is an enforced period of time between a dispute game's resolution and users being able to finalize withdrawals that were proven against its root claim. When the DisputeGame resolves globally, it stores the timestamp. The portal's finalizeWithdrawalTransaction function asserts that DISPUTE_GAME_FINALITY_DELAY_SECONDS have passed since the resolution timestamp before allowing any withdrawals proven against the dispute game to be finalized. Because the FaultDisputeGame is a trusted implementation set by the owner of the DisputeGameFactory, it is safe to trust that this value is honestly set.

Blacklisting DisputeGames

A new method is added to assign DisputeGames in the disputeGameBlacklist mapping mentioned in "State Layout", in the event that a dispute game is detected to have resolved incorrectly. The only actor who may call this function is the Guardian role.

Blacklisting a dispute game means that no withdrawals proven against it will be allowed to finalize (per the "Dispute Game Blacklist" invariant), and they must re-prove against a new dispute game that resolves correctly. The Portal's guardian role is obligated to blacklist any dispute games that it deems to have resolved incorrectly. Withdrawals proven against a blacklisted dispute game are not prevented from re-proving or being finalized in the future.

Blacklisting a full GameType

In the event of a catastrophic failure, we can upgrade the OptimismPortal proxy to an implementation with a different RESPECTED_GAME_TYPE. All pending withdrawals that reference a different game type will not be allowed to finalize and must re-prove, due to the "Trusted GameType" invariant. This should generally be avoided, but allows for a blanket blacklist of pending withdrawals corresponding to the current RESPECTED_GAME_TYPE. Depending on if we're okay with the tradeoffs, this also may be the most efficient way to upgrade the dispute game in the future.

Proxy Upgrade

Upgrading the OptimismPortal proxy to an implementation that follows this specification will invalidate all pending withdrawals. This means that all users with pending withdrawals will need to re-prove their withdrawals against an output proposal submitted in the form of a DisputeGame.

Permissioned FaultDisputeGame

As a fallback to permissioned proposals, a child contract of the FaultDisputeGame will be created that has 2 new roles: the PROPOSER and a CHALLENGER (or set of challengers). Each interaction (move [attack / defend], step, resolve / resolveClaim, addLocalData, etc.) will be permissioned to the CHALLENGER key, and the initialize function will be permissioned to the PROPOSER key.

In the event that we'd like to switch back to permissioned proposals, we can change the RESPECTED_GAME_TYPE in the OptimismPortal to a deployment of the PermissionedFaultDisputeGame.

Roles - PermissionedDisputeGame

• PROPOSER - Actor that can create a PermissionedFaultDisputeGame and participate in the games they've created.
• CHALLENGER - Actor(s) that can participate in a PermissionedFaultDisputeGame.

Modifications

State Layout

2 new immutables:

/// @notice The `PROPOSER` role.
address public immutable PROPOSER;

/// @notice The `CHALLENGER` role.
address public immutable CHALLENGER;

Functions

Every function that can mutate state should be overridden to add a check that either:

1. The msg.sender has the CHALLENGER role.
2. The msg.sender has the PROPOSER role.

If the msg.sender does not have either role, the function must revert.

The exception is the initialize function, which may only be called if the tx.origin is the PROPOSER role.

Precompiles

Table of Contents

• Overview
• P256VERIFY

Overview

Precompiled contracts exist on OP-Stack chains at predefined addresses. They are similar to predeploys but are implemented as native code in the EVM as opposed to bytecode. Precompiles are used for computationally expensive operations, that would be cost prohibitive to implement in Solidity. Where possible predeploys are preferred, as precompiles must be implemented in every execution client.

OP-Stack chains contain the standard Ethereum precompiles as well as a small number of additional precompiles. The following table lists each of the additional precompiles. The system version indicates when the precompile was introduced.

P256VERIFY

The P256VERIFY precompile performs signature verification for the secp256r1 elliptic curve. This curve has widespread adoption. It's used by Passkeys, Apple Secure Enclave and many other systems.

It is specified as part of RIP-7212 and was added to the OP-Stack protocol in the Fjord release. The op-geth implementation is here.

Address: 0x0000000000000000000000000000000000000100

Predeploys

Table of Contents

• Overview
• LegacyMessagePasser
• L2ToL1MessagePasser
• DeployerWhitelist
• LegacyERC20ETH
• WETH9
• L2CrossDomainMessenger
• L2StandardBridge
• L1BlockNumber
• GasPriceOracle
• L1Block
• ProxyAdmin
• SequencerFeeVault
• OptimismMintableERC20Factory
• OptimismMintableERC721Factory
• BaseFeeVault
• L1FeeVault
• SchemaRegistry
• EAS
• Beacon Block Root

Overview

Predeployed smart contracts exist on Optimism at predetermined addresses in the genesis state. They are similar to precompiles but instead run directly in the EVM instead of running native code outside of the EVM.

Predeploys are used instead of precompiles to make it easier for multiclient implementations as well as allowing for more integration with hardhat/foundry network forking.

Predeploy addresses exist in a prefixed namespace 0x4200000000000000000000000000000000000xxx. Proxies are set at the first 2048 addresses in the namespace, except for the addresses reserved for the GovernanceToken and WETH predeploys.

The LegacyERC20ETH predeploy lives at a special address 0xDeadDeAddeAddEAddeadDEaDDEAdDeaDDeAD0000 and there is no proxy deployed at that account.

The following table includes each of the predeploys. The system version indicates when the predeploy was introduced. The possible values are Legacy or Bedrock or Canyon. Deprecated contracts should not be used.

LegacyMessagePasser

Implementation

Address: 0x4200000000000000000000000000000000000000

The LegacyMessagePasser contract stores commitments to withdrawal transactions before the Bedrock upgrade. A merkle proof to a particular storage slot that commits to the withdrawal transaction is used as part of the withdrawing transaction on L1. The expected account that includes the storage slot is hardcoded into the L1 logic. After the bedrock upgrade, the L2ToL1MessagePasser is used instead. Finalizing withdrawals from this contract will no longer be supported after the Bedrock and is only left to allow for alternative bridges that may depend on it. This contract does not forward calls to the L2ToL1MessagePasser and calling it is considered a no-op in context of doing withdrawals through the CrossDomainMessenger system.

Any pending withdrawals that have not been finalized are migrated to the L2ToL1MessagePasser as part of the upgrade so that they can still be finalized.

L2ToL1MessagePasser

Implementation

Address: 0x4200000000000000000000000000000000000016

The L2ToL1MessagePasser stores commitments to withdrawal transactions. When a user is submitting the withdrawing transaction on L1, they provide a proof that the transaction that they withdrew on L2 is in the sentMessages mapping of this contract.

Any withdrawn ETH accumulates into this contract on L2 and can be permissionlessly removed from the L2 supply by calling the burn() function.

DeployerWhitelist

Implementation

Address: 0x4200000000000000000000000000000000000002

The DeployerWhitelist is a predeploy that was used to provide additional safety during the initial phases of Optimism. It previously defined the accounts that are allowed to deploy contracts to the network.

Arbitrary contract deployment was subsequently enabled and it is not possible to turn off. In the legacy system, this contract was hooked into CREATE and CREATE2 to ensure that the deployer was allowlisted.

In the Bedrock system, this contract will no longer be used as part of the CREATE codepath.

This contract is deprecated and its usage should be avoided.

LegacyERC20ETH

Implementation

Address: 0xDeadDeAddeAddEAddeadDEaDDEAdDeaDDeAD0000

The LegacyERC20ETH predeploy represents all ether in the system before the Bedrock upgrade. All ETH was represented as an ERC20 token and users could opt into the ERC20 interface or the native ETH interface.

The upgrade to Bedrock migrates all ether out of this contract and moves it to its native representation. All of the stateful methods in this contract will revert after the Bedrock upgrade.

This contract is deprecated and its usage should be avoided.

WETH9

Implementation

Address: 0x4200000000000000000000000000000000000006

WETH9 is the standard implementation of Wrapped Ether on Optimism. It is a commonly used contract and is placed as a predeploy so that it is at a deterministic address across Optimism based networks.

L2CrossDomainMessenger

Implementation

Address: 0x4200000000000000000000000000000000000007

The L2CrossDomainMessenger gives a higher level API for sending cross domain messages compared to directly calling the L2ToL1MessagePasser. It maintains a mapping of L1 messages that have been relayed to L2 to prevent replay attacks and also allows for replayability if the L1 to L2 transaction reverts on L2.

Any calls to the L1CrossDomainMessenger on L1 are serialized such that they go through the L2CrossDomainMessenger on L2.

The relayMessage function executes a transaction from the remote domain while the sendMessage function sends a transaction to be executed on the remote domain through the remote domain's relayMessage function.

L2StandardBridge

Implementation

Address: 0x4200000000000000000000000000000000000010

The L2StandardBridge is a higher level API built on top of the L2CrossDomainMessenger that gives a standard interface for sending ETH or ERC20 tokens across domains.

To deposit a token from L1 to L2, the L1StandardBridge locks the token and sends a cross domain message to the L2StandardBridge which then mints the token to the specified account.

To withdraw a token from L2 to L1, the user will burn the token on L2 and the L2StandardBridge will send a message to the L1StandardBridge which will unlock the underlying token and transfer it to the specified account.

The OptimismMintableERC20Factory can be used to create an ERC20 token contract on a remote domain that maps to an ERC20 token contract on the local domain where tokens can be deposited to the remote domain. It deploys an OptimismMintableERC20 which has the interface that works with the StandardBridge.

This contract can also be deployed on L1 to allow for L2 native tokens to be withdrawn to L1.

L1BlockNumber

Implementation

Address: 0x4200000000000000000000000000000000000013

The L1BlockNumber returns the last known L1 block number. This contract was introduced in the legacy system and should be backwards compatible by calling out to the L1Block contract under the hood.

It is recommended to use the L1Block contract for getting information about L1 on L2.

GasPriceOracle

Implementation

Address: 0x420000000000000000000000000000000000000F

In the legacy system, the GasPriceOracle was a permissioned contract that was pushed the L1 base fee and the L2 gas price by an offchain actor. The offchain actor observes the L1 blockheaders to get the L1 base fee as well as the gas usage on L2 to compute what the L2 gas price should be based on a congestion control algorithm.

After Bedrock, the GasPriceOracle is no longer a permissioned contract and only exists to preserve the API for offchain gas estimation. The function getL1Fee(bytes) accepts an unsigned RLP transaction and will return the L1 portion of the fee. This fee pays for using L1 as a data availability layer and should be added to the L2 portion of the fee, which pays for execution, to compute the total transaction fee.

The values used to compute the L1 portion of the fee prior to the Ecotone upgrade are:

• scalar
• overhead
• decimals

After the Bedrock upgrade, these values are instead managed by the SystemConfig contract on L1. The scalar and overhead values are sent to the L1Block contract each block and the decimals value has been hardcoded to 6.

Following the Ecotone upgrade, the values used for L1 fee computation are:

• baseFeeScalar
• blobBaseFeeScalar
• decimals

These new scalar values are managed by the SystemConfig contract on the L1 by introducing a backwards compatible versioned encoding scheme of its scalars storage slot. The decimals remains hardcoded to 6, and the overhead value is ignored.

L1Block

Implementation

Address: 0x4200000000000000000000000000000000000015

The L1Block was introduced in Bedrock and is responsible for maintaining L1 context in L2. This allows for L1 state to be accessed in L2.

ProxyAdmin

ProxyAdmin Address: 0x4200000000000000000000000000000000000018

The ProxyAdmin is the owner of all of the proxy contracts set at the predeploys. It is itself behind a proxy. The owner of the ProxyAdmin will have the ability to upgrade any of the other predeploy contracts.

SequencerFeeVault

Implementation

Address: 0x4200000000000000000000000000000000000011

The SequencerFeeVault accumulates any transaction priority fee and is the value of block.coinbase. When enough fees accumulate in this account, they can be withdrawn to an immutable L1 address.

To change the L1 address that fees are withdrawn to, the contract must be upgraded by changing its proxy's implementation key.

OptimismMintableERC20Factory

Implementation

Address: 0x4200000000000000000000000000000000000012

The OptimismMintableERC20Factory is responsible for creating ERC20 contracts on L2 that can be used for depositing native L1 tokens into. These ERC20 contracts can be created permissionlessly and implement the interface required by the StandardBridge to just work with deposits and withdrawals.

Each ERC20 contract that is created by the OptimismMintableERC20Factory allows for the L2StandardBridge to mint and burn tokens, depending on if the user is depositing from L1 to L2 or withdrawing from L2 to L1.

OptimismMintableERC721Factory

Implementation

Address: 0x4200000000000000000000000000000000000017

The OptimismMintableERC721Factory is responsible for creating ERC721 contracts on L2 that can be used for depositing native L1 NFTs into.

BaseFeeVault

Implementation

Address: 0x4200000000000000000000000000000000000019

The BaseFeeVault predeploy receives the base fees on L2. The base fee is not burnt on L2 like it is on L1. Once the contract has received a certain amount of fees, the ETH can be withdrawn to an immutable address on L1.

L1FeeVault

Implementation

Address: 0x420000000000000000000000000000000000001a

The L1FeeVault predeploy receives the L1 portion of the transaction fees. Once the contract has received a certain amount of fees, the ETH can be withdrawn to an immutable address on L1.

SchemaRegistry

Implementation

Address: 0x4200000000000000000000000000000000000020

The SchemaRegistry predeploy implements the global attestation schemas for the Ethereum Attestation Service protocol.

EAS

Implementation

Address: 0x4200000000000000000000000000000000000021

The EAS predeploy implements the Ethereum Attestation Service protocol.

Beacon Block Root

Address: 0x000F3df6D732807Ef1319fB7B8bB8522d0Beac02

The BeaconBlockRoot predeploy provides access to the L1 beacon block roots. This was added during the Ecotone network upgrade and is specified in EIP-4788.

Preinstalls

Table of Contents

• Overview
• Safe
• SafeL2
• MultiSend
• MultiSendCallOnly
• SafeSingletonFactory
• Multicall3
• Create2Deployer
• CreateX
• Arachnid's Deterministic Deployment Proxy
• Permit2
• ERC-4337 v0.6.0 EntryPoint
• ERC-4337 v0.6.0 SenderCreator
• ERC-4337 v0.7.0 EntryPoint
• ERC-4337 v0.7.0 SenderCreator

Overview

Preinstalled smart contracts exist on Optimism at predetermined addresses in the genesis state. They are similar to precompiles but instead run directly in the EVM instead of running native code outside of the EVM and are developed by third parties unaffiliated with the Optimism Collective.

These preinstalls are commonly deployed smart contracts that are being placed at genesis for convenience. It's important to note that these contracts do not have the same security guarantees as Predeployed smart contracts.

The following table includes each of the preinstalls.

Safe

Implementation

Address: 0x69f4D1788e39c87893C980c06EdF4b7f686e2938

A multisignature wallet with support for confirmations using signed messages based on ERC191. Differs from SafeL2 by not emitting events to save gas.

SafeL2

Implementation

Address: 0xfb1bffC9d739B8D520DaF37dF666da4C687191EA

A multisignature wallet with support for confirmations using signed messages based on ERC191. Differs from Safe by emitting events.

MultiSend

Implementation

Address: 0x998739BFdAAdde7C933B942a68053933098f9EDa

Allows to batch multiple transactions into one.

MultiSendCallOnly

Implementation

Address: 0xA1dabEF33b3B82c7814B6D82A79e50F4AC44102B

Allows to batch multiple transactions into one, but only calls.

SafeSingletonFactory

Implementation

Address: 0x914d7Fec6aaC8cd542e72Bca78B30650d45643d7

Singleton factory used by Safe-related contracts based on Arachnid's Deterministic Deployment Proxy.

The original library used a pre-signed transaction without a chain ID to allow deployment on different chains. Some chains do not allow such transactions to be submitted; therefore, this contract will provide the same factory that can be deployed via a pre-signed transaction that includes the chain ID. The key that is used to sign is controlled by the Safe team.

Multicall3

Implementation

Address: 0xcA11bde05977b3631167028862bE2a173976CA11

Multicall3 has two main use cases:

• Aggregate results from multiple contract reads into a single JSON-RPC request.
• Execute multiple state-changing calls in a single transaction.

Create2Deployer

Implementation

The create2Deployer is a nice Solidity wrapper around the CREATE2 opcode. It provides the following ABI.

    /**
     * @dev Deploys a contract using `CREATE2`. The address where the
     * contract will be deployed can be known in advance via {computeAddress}.
     *
     * The bytecode for a contract can be obtained from Solidity with
     * `type(contractName).creationCode`.
     *
     * Requirements:
     * - `bytecode` must not be empty.
     * - `salt` must have not been used for `bytecode` already.
     * - the factory must have a balance of at least `value`.
     * - if `value` is non-zero, `bytecode` must have a `payable` constructor.
     */
    function deploy(uint256 value, bytes32 salt, bytes memory code) public;
    /**
     * @dev Deployment of the {ERC1820Implementer}.
     * Further information: https://eips.ethereum.org/EIPS/eip-1820
     */
    function deployERC1820Implementer(uint256 value, bytes32 salt);
    /**
     * @dev Returns the address where a contract will be stored if deployed via {deploy}.
     * Any change in the `bytecodeHash` or `salt` will result in a new destination address.
     */
    function computeAddress(bytes32 salt, bytes32 codeHash) public view returns (address);
    /**
     * @dev Returns the address where a contract will be stored if deployed via {deploy} from a
     * contract located at `deployer`. If `deployer` is this contract's address, returns the
     * same value as {computeAddress}.
     */
    function computeAddressWithDeployer(
        bytes32 salt,
        bytes32 codeHash,
        address deployer
    ) public pure returns (address);

Address: 0x13b0D85CcB8bf860b6b79AF3029fCA081AE9beF2

When Canyon activates, the contract code at 0x13b0D85CcB8bf860b6b79AF3029fCA081AE9beF2 is set to 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.

CreateX

Implementation

Address: 0xba5Ed099633D3B313e4D5F7bdc1305d3c28ba5Ed

CreateX introduces additional logic for deploying contracts using CREATE, CREATE2 and CREATE3. It adds salt protection for sender and chainID and includes a set of helper functions.

The keccak256 of the CreateX bytecode is 0xbd8a7ea8cfca7b4e5f5041d7d4b17bc317c5ce42cfbc42066a00cf26b43eb53f.

Arachnid's Deterministic Deployment Proxy

Implementation

Address: 0x4e59b44847b379578588920cA78FbF26c0B4956C

This contract can deploy other contracts with a deterministic address on any chain using CREATE2. The CREATE2 call will deploy a contract (like CREATE opcode) but instead of the address being keccak256(rlp([deployer_address, nonce])) it instead uses the hash of the contract's bytecode and a salt. This means that a given deployer address will deploy the same code to the same address no matter when or where they issue the deployment. The deployer is deployed with a one-time-use account, so no matter what chain the deployer is on, its address will always be the same. This means the only variables in determining the address of your contract are its bytecode hash and the provided salt.

Between the use of CREATE2 opcode and the one-time-use account for the deployer, this contracts ensures that a given contract will exist at the exact same address on every chain, but without having to use the same gas pricing or limits every time.

Permit2

Implementation

Address: 0x000000000022D473030F116dDEE9F6B43aC78BA3

Permit2 introduces a low-overhead, next-generation token approval/meta-tx system to make token approvals easier, more secure, and more consistent across applications.

ERC-4337 v0.6.0 EntryPoint

Implementation

Address: 0x5FF137D4b0FDCD49DcA30c7CF57E578a026d2789

This contract verifies and executes the bundles of ERC-4337 v0.6.0 UserOperations sent to it.

ERC-4337 v0.6.0 SenderCreator

Implementation

Address: 0x7fc98430eaedbb6070b35b39d798725049088348

Helper contract for EntryPoint v0.6.0, to call userOp.initCode from a "neutral" address, which is explicitly not EntryPoint itself.

ERC-4337 v0.7.0 EntryPoint

Implementation

Address: 0x0000000071727De22E5E9d8BAf0edAc6f37da032

This contract verifies and executes the bundles of ERC-4337 v0.7.0 UserOperations sent to it.

ERC-4337 v0.7.0 SenderCreator

Implementation

Address: 0xEFC2c1444eBCC4Db75e7613d20C6a62fF67A167C

Helper contract for EntryPoint v0.7.0, to call userOp.initCode from a "neutral" address, which is explicitly not EntryPoint itself.

Superchain Configuration

Table of Contents

• Overview
• Configurable values
• Configuration data flow
  • Pausability
    • Paused identifiers
    • Scope of pausability

Overview

The SuperchainConfig contract is used to manage global configuration values for multiple OP Chains within a single Superchain network.

Configurable values

Configurability of the Superchain is currently limited to two values:

The SuperchainConfig contract manages the following configuration values:

• PAUSED_SLOT: A boolean value indicating whether the Superchain is paused.
• GUARDIAN_SLOT: The address of the guardian, which can pause and unpause the system.

Configuration data flow

All contracts which read from the SuperchainConfig contract hold its address as storage values in the proxy account, and call directly to it when reading configuration data.

flowchart TD
StandardBridge --> SuperchainConfig
L1ERC721Bridge --> SuperchainConfig
L1CrossDomainMessenger --> SuperchainConfig
OptimismPortal --> SuperchainConfig

Pausability

The Superchain pause feature is a safety mechanism designed to temporarily halt withdrawals from the system in an emergency situation. The Guardian role is authorized to pause and unpause the system.

Paused identifiers

When the system is paused the Paused(string identifier) event is emitted. This allows for the caller to provide additional information to be used during incident response.

Scope of pausability

The pause applies specifically to withdrawals of assets from the L1 bridge contracts. The L2 bridge contracts are not pausable, on the basis that issues on L2 can be addressed more easily by a hard fork in the consensus layer.

When the Pause is activated, the following methods are disabled:

1. OptimismPortal.proveWithdrawalTransaction()
2. OptimismPortal.finalizeWithdrawalTransaction()
3. L1CrossDomainMessenger.relayMessage()
4. StandardBridge.finalizeBridgeERC20()
5. StandardBridge.finalizeBridgeETH()
6. L1ERC721Bridge.finalizeBridgeERC721()

Superchain Upgrades

Table of Contents

• Overview
• Protocol Version
  • Protocol Version Format
    • Build identifier
    • Major versions
    • Minor versions
    • Patch versions
    • Pre-releases
  • Protocol Version Exposure
• Superchain Target
  • Superchain Version signaling
  • ProtocolVersions L1 contract
• Activation rules
  • L2 Block-number based activation (deprecated)
  • L2 Block-timestamp based activation
• OP-Stack Protocol versions
• Post-Bedrock Network upgrades
  • Activation Timestamps

Overview

Superchain upgrades, also known as forks or hardforks, implement consensus-breaking changes.

A Superchain upgrade requires the node software to support up to a given Protocol Version. The version indicates support, the upgrade indicates the activation of new functionality.

This document lists the protocol versions of the OP-Stack, starting at the Bedrock upgrade, as well as the default Superchain Targets.

Activation rule parameters of network upgrades are configured as part of the Superchain Target specification: chains following the same Superchain Target upgrade synchronously.

Protocol Version

The Protocol Version documents the progression of the total set of canonical OP-Stack specifications. Components of the OP-Stack implement the subset of their respective protocol component domain, up to a given Protocol Version of the OP-Stack.

OP-Stack mods, i.e. non-canonical extensions to the OP-Stack, are not included in the versioning of the Protocol. Instead, mods must specify which upstream Protocol Version they are based on and where breaking changes are made. This ensures tooling of the OP-Stack can be shared and collaborated on with OP-Stack mods.

The Protocol Version is NOT a hardfork identifier, but rather indicates software-support for a well-defined set of features introduced in past and future hardforks, not the activation of said hardforks.

Changes that can be included in prospective Protocol Versions may be included in the specifications as proposals, with explicit notice of the Protocol Version they are based on. This enables an iterative integration process into the canonical set of specifications, but does not guarantee the proposed specifications become canonical.

Note that the Protocol Version only applies to the Protocol specifications with the Superchain Targets specified within. This versioning is independent of the Semver versioning used in OP Stack smart-contracts, and the Semver-versioned reference software of the OP-Stack.

Protocol Version Format

The Protocol Version is Semver-compatible. It is encoded as a single 32 bytes long <protocol version>. The version must be encoded as 32 bytes of DATA in JSON RPC usage.

The encoding is typed, to ensure future-compatibility.

<protocol version> ::= <version-type><typed-payload>
<version-type> ::= <uint8>
<typed-payload> ::= <31 bytes>

version-type 0:

<reserved><build><major><minor><patch><pre-release>
<reserved> ::= <7 zeroed bytes>
<build> ::= <8 bytes>
<major> ::= <big-endian uint32>
<minor> ::= <big-endian uint32>
<patch> ::= <big-endian uint32>
<pre-release> ::= <big-endian uint32>

The <reserved> bytes of the Protocol Version are reserved for future extensions.

Protocol versions with a different <version-type> should not be compared directly.

Build identifier

The <build> identifier, as defined by Semver, is ignored when determining version precedence. The <build> must be non-zero to apply to the protocol version.

Modifications of the OP-Stack should define a <build> to distinguish from the canonical protocol feature-set. Changes to the <build> may be encoded in the <build> itself to stay aligned with the upstream protocol. The major/minor/patch versions should align with that of the upstream protocol that the modifications are based on. Users of the protocol can choose to implement custom support for the alternative <build>, but may work out of the box if the major features are consistent with that of the upstream protocol version.

The 8 byte <build> identifier may be presented as a string for human readability if the contents are alpha-numeric, including - and ., as outlined in the Semver format specs. Trailing 0 bytes can be used for padding. It may be presented as 0x-prefixed hex string otherwise.

Major versions

Major version changes indicate support for new consensus-breaking functionality. Major versions should retain support for functionality of previous major versions for syncing/indexing of historical chain data. Implementations may drop support for previous Major versions, when there are viable alternatives, e.g. l2geth for pre-Bedrock data.

Minor versions

Minor version changes indicate support for backward compatible extensions, including backward-compatible additions to the set of chains in a Superchain Target. Backward-compatibility is defined by the requirement for existing end-users to upgrade nodes and tools or not. Minor version changes may also include optional offchain functionality, such as additional syncing protocols.

Patch versions

Patch version changes indicate backward compatible bug fixes and improvements.

Pre-releases

Pre-releases of the protocol are proposals: these are not stable targets for production usage. A pre-release might not satisfy the intended compatibility requirements as denoted by its associated normal version. The <pre-release> must be non-zero to apply to the protocol version. The <pre-release> 0-value is reserved for non-prereleases, i.e. v3.1.0 is higher than v3.1.0-1.

Node-software may support a pre-release, but must not activate any protocol changes without the user explicitly opting in through the means of a feature-flag or configuration change.

A pre-release is not an official version and is meant for protocol developers to communicate an experimental changeset before the changeset is reviewed by governance. Pre-releases are subject to change.

Protocol Version Exposure

The Protocol Version is not exposed to the application-layer environment: hardforks already expose the change of functionality upon activation as required, and the Protocol Version is meant for offchain usage only. The protocol version indicates support rather than activation of functionality. There is one exception however: signaling by onchain components to offchain components. More about this in Superchain Version signaling.

Superchain Target

Changes to the L2 state-transition function are transitioned into deterministically across all nodes through an activation rule.

Changes to L1 smart-contracts must be compatible with the latest activated L2 functionality, and are executed through L1 contract-upgrades.

A Superchain Target defines a set of activation rules and L1 contract upgrades shared between OP-Stack chains, to upgrade the chains collectively.

Superchain Version signaling

Each Superchain Target tracks the protocol changes, and signals the recommended and required Protocol Version ahead of activation of new breaking functionality.

• recommended: a signal in advance of a network upgrade, to alert users of the protocol change to be prepared for. Node software is recommended to signal the recommendation to users through logging and metrics.
• required: a signal shortly in advance of a breaking network upgrade, to alert users of breaking changes. Users may opt in to elevated alerts or preventive measures, to ensure consistency with the upgrade.

Signaling is done through a L1 smart-contract that is monitored by the OP-Stack software. Not all components of the OP-Stack are required to directly monitor L1 however: cross-component APIs like the Engine API may be used to forward the Protocol Version signals, to keep components encapsulated from L1. See engine_signalOPStackVersionV1.

ProtocolVersions L1 contract

The ProtocolVersions contract on L1 enables L2 nodes to pick up on superchain protocol version signals.

The interface is:

• Required storage slot: bytes32(uint256(keccak256("protocolversion.required")) - 1)
• Recommended storage slot: bytes32(uint256(keccak256("protocolversion.recommended")) - 1)
• Required getter: required() returns ProtocolVersion
• Recommended getter recommended() returns ProtocolVersion
• Version updates also emit a typed event: event ConfigUpdate(uint256 indexed version, UpdateType indexed updateType, bytes data)

Activation rules

The below L2-block based activation rules may be applied in two contexts:

• The rollup node, specified through the rollup configuration (known as rollup.json), referencing L2 blocks (or block input-attributes) that pass through the derivation pipeline.
• The execution engine, specified through the chain configuration (known as the config part of genesis.json), referencing blocks or input-attributes that are part of, or applied to, the L2 chain.

For both types of configurations, some activation parameters may apply to all chains within the superchain, and are then retrieved from the superchain target configuration.

L2 Block-number based activation (deprecated)

Activation rule: upgradeNumber != null && block.number >= upgradeNumber

Starting at, and including, the L2 block with block.number >= upgradeNumber, the upgrade rules apply. If the upgrade block-number upgradeNumber is not specified in the configuration, the upgrade is ignored.

This block number based method has commonly been used in L1 up until the Bellatrix/Paris upgrade, a.k.a. The Merge, which was upgraded through special rules.

This method is not superchain-compatible, as the activation-parameter is chain-specific (different chains may have different block-heights at the same moment in time).

This applies to the L2 block number, not to the L1-origin block number. This means that an L2 upgrade may be inactive, and then active, without changing the L1-origin.

L2 Block-timestamp based activation

Activation rule: upgradeTime != null && block.timestamp >= upgradeTime

Starting at, and including, the L2 block with block.timestamp >= upgradeTime, the upgrade rules apply. If the upgrade block-timestamp upgradeTime is not specified in the configuration, the upgrade is ignored.

This is the preferred superchain upgrade activation-parameter type: it is synchronous between all L2 chains and compatible with post-Merge timestamp-based chain upgrades in L1.

This applies to the L2 block timestamp, not to the L1-origin block timestamp. This means that an L2 upgrade may be inactive, and then active, without changing the L1-origin.

This timestamp based method has become the default on L1 after the Bellatrix/Paris upgrade, a.k.a. The Merge, because it can be planned in accordance with beacon-chain epochs and slots.

Note that the L2 version is not limited to timestamps that match L1 beacon-chain slots or epochs. A timestamp may be chosen to be synchronous with a specific slot or epoch on L1, but the matching L1-origin information may not be present at the time of activation on L2.

OP-Stack Protocol versions

• v1.0.0: 2021 Jan 16th - Mainnet Soft Launch, based on OVM. (announcement)
• v1.1.0: 2021 Aug 19th - Community launch. (announcement)
• v2.0.0: 2021 Nov 12th - the EVM-Equivalence update, also known as OVM 2.0 and chain regenesis. (announcement)
• v2.1.0: 2022 May 31st - Optimism Collective. (announcement).
• v3.0.0-1: 2023 Jan 13th - Bedrock pre-release, deployed on OP-Goerli, and later Base-Goerli.
• v3.0.0: 2023 Jun 6th - Bedrock, including the Regolith hardfork improvements, first deployed on OP-Mainnet.
• v4.0.0: 2024 Jan 11th - Canyon network upgrade (Shapella). Governance Proposal.
• v5.0.0: 2024 Feb 22nd - Delta network upgrade (Span Batches). Governance Proposal.
• v6.0.0: 2024 Mar 14th - Ecotone network upgrade (4844 Blob Batches + Cancun). Governance Proposal.
• v7.0.0: 2024 Jul 10th - Fjord network upgrade (RIP-7212 precompile + FastLZ cost fn + Brotli compression). Governance Proposal.
• v8.0.0: 2024 Sep 11th - Granite network upgrade (Limit ecpairing input size + Reduce Channel Timeout). Governance Proposal.

Post-Bedrock Network upgrades

Activation Timestamps

Governance approves all network upgrades & the time at which the upgrade activates. The approved governance proposal is the canonical document for the timestamp; however, the timestamps are replicated here for ease of use.

System Config

Table of Contents

• Overview
• System config contents (version 0)
  • batcherHash (bytes32)
  • Scalars
    • Pre-Ecotone scalar, overhead (uint256,uint256)
    • Ecotone scalar, overhead (uint256,uint256) change
  • gasLimit (uint64)
  • unsafeBlockSigner (address)
• Writing the system config
• Reading the system config

Overview

The SystemConfig is a contract on L1 that can emit rollup configuration changes as log events. The rollup block derivation process picks up on these log events and applies the changes.

System config contents (version 0)

Version 0 of the system configuration contract defines the following parameters:

batcherHash (bytes32)

A versioned hash of the current authorized batcher sender(s), to rotate keys as batch-submitter. The first byte identifies the version.

Version 0 embeds the current batch submitter ethereum address (bytes20) in the last 20 bytes of the versioned hash.

In the future this versioned hash may become a commitment to a more extensive configuration, to enable more extensive redundancy and/or rotation configurations.

Scalars

The L1 fee parameters, also known as Gas Price Oracle (GPO) parameters, are used to compute the L1 data fee applied to an L2 transaction. The specific parameters used depend on the upgrades that are active.

Fee parameter updates are signaled to L2 through the GAS_CONFIG log-event of the SystemConfig.

Pre-Ecotone scalar, overhead (uint256,uint256)

The overhead and scalar are consulted and passed to the L2 via L1 attribute info.

The values are interpreted as big-endian uint256.

Ecotone scalar, overhead (uint256,uint256) change

After Ecotone activation:

• The scalar attribute encodes additional scalar information, in a versioned encoding scheme.
• The overhead value is ignored: it does not affect the L2 state-transition output.

The scalar is encoded as big-endian uint256, interpreted as bytes32, and composed as following:

*Byte ranges are indicated with [ (inclusive) and ) (exclusive).

• 0: scalar-version byte
• [1, 32): depending scalar-version:
  • Scalar-version 0:
    • [1, 28): padding, should be zero.
    • [28, 32): big-endian uint32, encoding the L1-fee baseFeeScalar
    • This version implies the L1-fee blobBaseFeeScalar is set to 0.
    • In the event there are non-zero bytes in the padding area, baseFeeScalar must be set to MaxUint32.
    • This version is compatible with the pre-Ecotone scalar value (assuming a uint32 range).
  • Scalar-version 1:
    • [1, 24): padding, must be zero.
    • [24, 28): big-endian uint32, encoding the blobBaseFeeScalar
    • [28, 32): big-endian uint32, encoding the baseFeeScalar
    • This version is meant to configure the EIP-4844 blob fee component, once blobs are used for data-availability.
  • Other scalar-version values: unrecognized. OP-Stack forks are recommended to utilize the >= 128 scalar-version range and document their scalar encoding.

Invalid and unrecognized scalar event-data should be ignored, and the last valid configuration should continue to be utilized.

The baseFeeScalar and blobBaseFeeScalar are incorporated into the L2 through the Ecotone L1 attributes deposit transaction calldata.

Future upgrades of the SystemConfig contract may provide additional typed getters/setters for the versioned scalar information.

In Ecotone the existing setGasConfig function, and scalar and overhead getters, continue to function.

When the batch-submitter utilizes EIP-4844 blob data for data-availability it can adjust the scalars to accurately price the resources:

• baseFeeScalar to correspond to the share of the user-transaction (per byte) in the total regular L1 EVM gas usage consumed by the data-transaction of the batch-submitter. For blob transactions this is the fixed intrinsic gas cost of the L1 transaction.

• blobBaseFeeScalar to correspond to share of a user-transaction (per byte) in the total Blob data that is introduced by the data-transaction of the batch-submitter.

gasLimit (uint64)

The gas limit of the L2 blocks is configured through the system config. Changes to the L2 gas limit are fully applied in the first L2 block with the L1 origin that introduced the change, as opposed to the 1/1024 adjustments towards a target as seen in limit updates of L1 blocks.

The gas limit may not be set to a value larger than 200_000_000. This is to ensure that the L2 blocks are fault provable and of reasonable size to be processed by the client software. Over time, this value will be increased.

unsafeBlockSigner (address)

Blocks are gossiped around the p2p network before they are made available on L1. To prevent denial of service on the p2p layer, these unsafe blocks must be signed with a particular key to be accepted as "canonical" unsafe blocks. The address corresponding to this key is the unsafeBlockSigner. To ensure that its value can be fetched with a storage proof in a storage layout independent manner, it is stored at a special storage slot corresponding to keccak256("systemconfig.unsafeblocksigner").

Unlike the other values, the unsafeBlockSigner only operates on blockchain policy. It is not a consensus level parameter.

Writing the system config

The SystemConfig contract applies authentication to all writing contract functions, the configuration management can be configured to be any type of ethereum account or contract.

On a write, an event is emitted for the change to be picked up by the L2 system, and a copy of the new written configuration variable is retained in L1 state to read with L1 contracts.

Reading the system config

A rollup node initializes its derivation process by finding a starting point based on its past L2 chain:

• When started from L2 genesis, the initial system configuration is retrieved from the rollup chain configuration.
• When started from an existing L2 chain, a previously included L1 block is determined as derivation starting point, and the system config can thus be retrieved from the last L2 block that referenced the L1 block as L1 origin:
  • If the chain state precedes the Ecotone upgrade, batcherHash, overhead and scalar are retrieved from the L1 block info transaction. Otherwise, batcherHash, baseFeeScalar, and blobBaseFeeScalar are retrieved instead.
  • gasLimit is retrieved from the L2 block header.
  • other future variables may also be retrieved from other contents of the L2 block, such as the header.

After preparing the initial system configuration for the given L1 starting input, the system configuration is updated by processing all receipts from each new L1 block.

The contained log events are filtered and processed as follows:

• the log event contract address must match the rollup SystemConfig deployment
• the first log event topic must match the ABI hash of ConfigUpdate(uint256,uint8,bytes)
• the second topic determines the version. Unknown versions are critical derivation errors.
• the third topic determines the type of update. Unknown types are critical derivation errors.
• the remaining event data is opaque, encoded as ABI bytes (i.e. includes offset and length data), and encodes the configuration update. In version 0 the following types are supported:
  • type 0: batcherHash overwrite, as bytes32 payload.
  • type 1: Pre-Ecotone, overhead and scalar overwrite, as two packed uint256 entries. After Ecotone upgrade, overhead is ignored and scalar interpreted as a versioned encoding that updates baseFeeScalar and blobBaseFeeScalar.
  • type 2: gasLimit overwrite, as uint64 payload.
  • type 3: unsafeBlockSigner overwrite, as address payload.

Note that individual derivation stages may be processing different L1 blocks, and should thus maintain individual system configuration copies, and apply the event-based changes as the stage traverses to the next L1 block.

OP Stack Configurability

Table of Contents

• Consensus Parameters
  • Batch Inbox address
  • Batcher Hash
  • Chain ID
  • Proof Maturity Delay
  • Dispute Game Finality
  • Respected Game Type
  • Fault Game Max Depth
  • Fault Game Split Depth
  • Max Game Clock Duration
  • Game Clock Extension
  • Bond Withdrawal Delay
  • Minimum Large Preimage Proposal Size
  • Large Preimage Proposal Challenge Period
  • Fault Game Absolute Prestate
  • Fault Game Genesis Block
  • Fault Game Genesis Output Root
  • Fee Scalar
  • Gas Limit
  • Genesis state
  • L2 block time
  • Resource config
  • Sequencing window Size
  • Start block
  • Superchain target
  • Governance Token
  • Resource Config
• Policy Parameters
  • Data Availability Type
  • Batch submission frequency
  • Output frequency
• Admin Roles
  • L1 Proxy Admin
  • L1 ProxyAdmin owner
  • L2 Proxy Admin
  • L2 ProxyAdmin owner
  • System Config Owner
• Service Roles
  • Batch submitter address
  • Challenger address
  • Guardian address
  • Proposer address
  • Sequencer P2P / Unsafe head signer

When deploying the OP Stack software to a production environment, certain parameters about the protocol can be configured. These configurations can include a variety of parameters which affect the resulting properties of the blockspace in question.

There are four categories of OP Stack configuration options:

• Consensus Parameters: Parameters and properties of the chain that may be set at genesis and fixed for the lifetime of the chain, or may be changeable through a privileged account or protocol upgrade.
• Policy Parameters: Parameters of the chain that might be changed without breaking consensus. Consensus is enforced by the protocol, while policy parameters may be changeable within constraints imposed by consensus.
• Admin Roles: These roles can upgrade contracts, change role owners, or update protocol parameters. These are typically cold/multisig wallets not used directly in day to day operations.
• Service Roles: These roles are used to manage the day-to-day operations of the chain, and therefore are often hot wallets.

Each category also defines the standard configuration values for it's given parameters. Standard configuration is the set of requirements for an OP Stack chain to be considered a Standard Chain within the superchain. These requirements are currently a draft, pending governance approval.

The recommended way to deploy L1 contracts for an OP chain that meet the standard configuration will be with the OP Contracts Manager.

Consensus Parameters

Batch Inbox address

Description: L1 address where calldata/blobs are posted ( see Batcher Transaction).
Administrator: Static
Requirement: Current convention is versionByte || keccak256(bytes32(chainId))[:19], where || denotes concatenation, versionByte is 0x00, and chainId is a uint256.
Notes: It is recommended, but not required, to follow this convention.

Batcher Hash

Description: A versioned hash of the current authorized batcher sender(s).
Administrator: System Config Owner
Requirement: bytes32(uint256(uint160(batchSubmitterAddress)))
Notes: Batch Submitter address padded with zeros to fit 32 bytes.

Chain ID

Description: Unique ID of Chain used for TX signature validation.
Administrator: Static
Requirement: Foundation-approved, globally unique value ¹.
Notes: Foundation will ensure chains are responsible with their chain IDs until there's a governance process in place.

Proof Maturity Delay

Description: The length of time that must pass between proving and finalizing a withdrawal.
Administrator: L1 Proxy Admin
Requirement: 7 days
Notes: High security. Excessively safe upper bound that leaves enough time to consider social layer solutions to a hack if necessary. Allows enough time for other network participants to challenge the integrity of the corresponding output root.

Dispute Game Finality

Description: The amount of time given to the Guardian role to blacklist a resolved dispute game before any withdrawals proven against it can be finalized, in the case of a system failure.
Administrator: L1 Proxy Admin
Requirement: 3.5 days
Notes: High security. Allows enough time for the Guardian to blacklist games.

Respected Game Type

Description: The respected game type of the OptimismPortal. Determines the type of dispute games that can be used to finalize withdrawals.
Administrator: Guardian
Requirement: CANNON ( 0)
Notes: The game type may be changed to PERMISSIONED_CANNON ( 1) as a fallback to permissioned proposals, in the event of a failure in the Fault Proof system.

Fault Game Max Depth

Description: The maximum depth of fault dispute game trees.
Administrator: Static
Requirement: 73
Notes: Sufficiently large to ensure the fault proof VM execution trace fits within the number of leaf nodes.

Fault Game Split Depth

Description: The depth in fault dispute game trees after which claims correspond to VM state commitments instead of output root commitments.
Administrator: Static
Requirement: 30
Notes: Sufficiently large to ensure enough nodes at the split depth to represent all L2 blocks since the anchor state.

Max Game Clock Duration

Description: The maximum amount of time that may accumulate on a dispute game team's chess clock.
Administrator: Static
Requirement: 3.5 days
Notes: High security. Allows enough time for honest actors to counter invalid claims.

Game Clock Extension

Description: The flat credit that is given to a dispute game team's clock if their clock has less than CLOCK_EXTENSION seconds remaining.
Administrator: Static
Requirement: 3 hours
Notes: Allows enough time for honest actors to counter freeloader claims.

Bond Withdrawal Delay

Description: The length of time that must pass before dispute game bonds can be withdrawn.
Administrator: Static
Requirement: 7 days
Notes: High security. Allows enough time for the Guardian to recover funds from DelayedWETH if bonds were allocated incorrectly.

Minimum Large Preimage Proposal Size

Description: The minimum size of preimage allowed to be submitted via the PreimageOracle large preimage proposal process.
Administrator: Static
Requirement: 126000 bytes
Notes: Large enough to ensure posting the large preimage is expensive enough to act as a deterrent but small enough to be used for any preimage that is too large to be submitted in a single transaction.

Large Preimage Proposal Challenge Period

Description: The amount of time that large preimage proposals can be challenged before they can be published to the PreimageOracle
Administrator: Static
Requirement: 24 hours
Notes: High security. Allows enough time for honest actors to challenge invalid large preimage proposals.

Fault Game Absolute Prestate

Description: The VM state commitment to use as the starting point when executing the fault proof VM
Administrator: Static
Requirement: The state commitment of a governance approved op-program release.
Notes: The op-program version must have the rollup config and L2 genesis of the chain built in via the superchain registry.

Fault Game Genesis Block

Description: The L2 block number used as the initial anchor state for fault dispute games
Administrator: Static
Requirement: Block number of any finalized block between bedrock activation and enabling fault proofs. 0 for chains using fault proofs from genesis.

Fault Game Genesis Output Root

Description: The output root at the Fault Game Genesis Block
Administrator: Static
Requirement: The output root from the canonical chain at Fault game Genesis Block.

Fee Scalar

Description: Markup on transactions compared to the raw L1 data cost.
Administrator: System Config Owner
Requirement: Set such that Fee Margin is between 0 and 50%.

Gas Limit

Description: Gas limit of the L2 blocks is configured through the system config.
Administrator: System Config Owner
Requirement: No higher than 200_000_000 gas
Notes: Chain operators are driven to maintain a stable and reliable chain. When considering to change this value, careful deliberation is necessary.

Genesis state

Description: Initial state at chain genesis, including code and storage of predeploys (all L2 smart contracts). See Predeploy.
Administrator: Static
Requirement: Only standard predeploys and preinstalls, no additional state.
Notes: Homogeneity & standardization, ensures initial state is secure.

L2 block time

Description: Frequency with which blocks are produced as a result of derivation.
Administrator: L1 Proxy Admin
Requirement: 1 or 2 seconds
Notes: High security & interoperability compatibility requirement, until de-risked/solved at app layer.

Resource config

Description: Config for the EIP-1559 based curve used for the deposit gas market.
Administrator: L1 Proxy Admin
Requirement: See resource config table.
Notes: Constraints are imposed in code when setting the resource config.

Sequencing window Size

Description: Maximum allowed batch submission gap, after which L1 fallback is triggered in derivation.
Administrator: Static
Requirement: 3_600 base layer blocks (12 hours for an L2 on Ethereum, assuming 12 second L1 blocktime). e.g. 12 second blocks, $3600\ast 12seconds÷60\frac{seconds}{minute}÷60\frac{minute}{hour}=12hours$.
Notes: This is an important value for constraining the sequencer's ability to re-order transactions; higher values would pose a risk to user protections.

Start block

Description: Block at which the system config was initialized the first time.
Administrator: L1 Proxy Admin
Requirement: The block where the SystemConfig was initialized.
Notes: Simple clear restriction.

Superchain target

Description: Choice of cross-L2 configuration. May be omitted in isolated OP Stack deployments. Includes SuperchainConfig and ProtocolVersions contract addresses.
Administrator: Static
Requirement: Mainnet or Sepolia
Notes: A superchain target defines a set of layer 2 chains which share SuperchainConfig and ProtocolVersions contracts deployed on layer 1.

Governance Token

Description: OP token used for the Optimism Collective's Token House governance.
Administrator: n/a
Requirement: Disabled
Notes: Simple clear restriction.

Resource Config

Policy Parameters

Data Availability Type

Description: Batcher can currently be configured to use blobs or calldata ( See Data Availability Provider).
Administrator: Batch submitter address
Requirement: Ethereum (Blobs, Calldata)
Notes: Alt-DA is not yet supported for the standard configuration, but the sequencer can switch at-will between blob and calldata with no restriction, since both are L1 security.

Batch submission frequency

Description: Frequency with which batches are submitted to L1 ( see Batcher Transaction).
Administrator: Batch submitter address
Requirement: 1_800 base layer blocks (6 hours for an L2 on Ethereum, assuming 12 second L1 blocktime) or lower
Notes: Batcher needs to fully submit each batch within the sequencing window, so this leaves buffer to account for L1 network congestion and the amount of data the batcher would need to post.

Output frequency

Description: Frequency with which output roots are submitted to L1.
Administrator: L1 Proxy Admin
Requirement: 43_200 L2 Blocks (24 hours for an L2 on Ethereum, assuming 2 second L2 blocktime) or lower e.g. 2 second blocks, $43200\ast 2seconds÷60\frac{seconds}{minute}÷60\frac{minute}{hour}=24hours$.
Notes: Deprecated once fault proofs are implemented. This value cannot be 0.