Skip to content

Latest commit

 

History

History
932 lines (692 loc) · 52 KB

derivation.md

File metadata and controls

932 lines (692 loc) · 52 KB
Raw

L2 Chain Derivation Specification

Table of Contents

Overview

Note the following assumes a single sequencer. 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 responsibility of the rollup node, both in syncer 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 is mapped to an L2 sequencing epoch comprising multiple L2 blocks. The epoch number is defined to be equal to the corresponding L1 block number.

To derive the L2 blocks in an epoch E, we need the following inputs:

  • The L1 sequencing window for epoch E: the L1 blocks in the range [E, E + PWS) where PWS is the sequencing window size (note that this means that epochs are overlapping). In particular, we need:
    • The batcher transactions included in the sequencing window. These allow us to reconstruct sequencer-batches containing the transactions to include in L2 blocks (each batch maps to a single L2 block).
      • Note that it is impossible to have a batcher transaction containing a batch relative to epoch E on L1 block E, as the batch must contain the hash of L1 block E.
    • The deposits made in L1 block E (in the form of events emitted by the deposit contract).
    • The L1 block attributes from L1 block E (to derive the L1 attributes deposited transaction).
  • The state of the L2 chain after the last L2 block of epoch E - 1, or — if epoch E - 1 does not exist — the L2 genesis state.

To derive the whole L2 chain from scratch, we simply start with the L2 genesis state, and the L2 chain inception as first epoch, then process all sequencing windows in order. Refer to the Architecture section for more information on how we implement this in practice.

Each epoch may contain a variable number of L2 blocks (one every l2_block_time, 2s on Kroma), at the discretion of the sequencer, but subject to the following constraints for each block:

  • min_l2_timestamp <= block.timestamp <= max_l2_timestamp, where
    • all these values are denominated in seconds
    • min_l2_timestamp = l1_timestamp
      • This ensures that the L2 timestamp is not behind the L1 origin timestamp.
    • block.timestamp = prev_l2_timestamp + l2_block_time
      • prev_l2_timestamp is the timestamp of the last L2 block of the previous epoch
      • l2_block_time is a configurable parameter of the time between L2 blocks (on Kroma, 2s)
    • max_l2_timestamp = max(l1_timestamp + max_sequencer_drift, min_l2_timestamp + l2_block_time)
      • l1_timestamp is the timestamp of the L1 block associated with the L2 block's epoch
      • max_sequencer_drift is the most a sequencer is allowed to get ahead of L1

Put together, these constraints mean that there must be an L2 block every l2_block_time seconds, and that the timestamp for the first L2 block of an epoch must never fall behind the timestamp of the L1 block matching the epoch.

Post-merge, Ethereum has a fixed block time of 12s (though some slots can be skipped). It is thus expected that with a 2-second L2 block time, most of the time, each epoch will contain 12/2 = 6 L2 blocks. The sequencer can however lengthen or shorten epochs (subject to above constraints). The rationale is to maintain liveness in case of either a skipped slot on L1, or a temporary loss of connection to L1 — which requires longer epochs. Shorter epochs are then required to avoid L2 timestamps drifting further and further ahead of L1.

Note that min_l2_timestamp + l2_block_time ensures that a new L2 batch can always be processed, even if the max_sequencer_drift is exceeded. However, when exceeding the max_sequencer_drift, progression to the next L1 origin is enforced, with an exception to ensure the minimum timestamp bound (based on this next L1 origin) can be met in the next L2 batch, and len(batch.transactions) == 0 continues to be enforced while the max_sequencer_drift is exceeded. See [Batch Queue] for more details.

Eager Block Derivation

In practice, it is often not necessary to wait for a full sequencing window of L1 blocks in order to start deriving the L2 blocks in an epoch. Indeed, as long as we are able to reconstruct sequential batches, we can start deriving the corresponding L2 blocks. We call this eager block derivation.

However, in the very worst case, we can only reconstruct the batch for the first L2 block in the epoch by reading the last L1 block of the sequencing window. This happens when some data for that batch is included in the last L1 block of the window. In that case, not only can we not derive the first L2 block in the epoch, we also cannot derive any further L2 block in the epoch until then, as they need the state that results from applying the epoch's first L2 block. (Note that this only applies to block derivation. Batches can still be derived and tentatively queued, we just won't be able to create blocks from them.)

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 a validator applies a state transition incorrectly and posts an output root, this output root will be incorrect. The incorrect output root, which will be challenged by a ZK fault proof, will then be 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 utilisation in a batcher transaction: for instance it allows us to pack the final (small) frame of a window with large frames from the next window.

In the future, this channel identification feature also allows the batcher to employ multiple signers (private keys) to submit one or multiple channels in parallel (1).

(1) This helps alleviate issues where, because of transaction nonce values affecting the L2 tx-pool and thus inclusion: multiple transactions made by the same signer are stuck waiting on the inclusion of a previous transaction.

Also note that we use a streaming compression scheme, and we do not need to know how many blocks 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.

batch derivation chain diagram

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 belongs to epoch 99.

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

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).

version_byte rollup_payload
0 frame ... (one or more frames, concatenated)

Unknown versions make the batcher transaction invalid (it must be ignored by the rollup node). All frames in a batcher transaction must be parsable. 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 as channel_encoding, defined as:

rlp_batches = []
for batch in batches:
    rlp_batches.append(batch)
channel_encoding = compress(rlp_batches)

where:

  • batches is the input, a sequence of batches byte-encoded as per the next section ("Batch Encoding")
  • rlp_batches is the concatenation of the RLP-encoded batches
  • compress is a function performing compression, using the ZLIB algorithm (as specified in RFC-1950) with no dictionary
  • channel_encoding is the compressed version of rlp_batches

When decompressing a channel, we limit the amount of decompressed data to MAX_RLP_BYTES_PER_CHANNEL (currently 10,000,000 bytes), 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 ven 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:

batch_version content
0 rlp_encode([parent_hash, epoch_number, epoch_hash, timestamp, transaction_list])

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.

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

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 validator 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 it. By default, the rollup operates on calldata retrieved from batcher transactions in the block, for each transaction:

  • The receiver must be the configured batcher inbox address.
  • The sender must match the batcher address loaded from the system config matching the L1 block of the data.

Each data-transaction is versioned and contains a series of channel frames to be read by the Frame Queue, see Batch Submission Wire Format.

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 of 100,000,000 bytes.

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

The channel-bank can only output data from the first opened channel.

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

Once the first opened channel, if any, is not timed-out and is ready, then it is read and removed from the channel-bank.

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 yet 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 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 + sequencer_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 + sequencer_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 syncer 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:

  • engine_forkchoiceUpdatedV1 — 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_getPayloadV1 — retrieves a previously requested execution payload build.
  • engine_newPayloadV1 — executes an execution payload to create a block.

The execution payload is an object of type ExecutionPayloadV1.

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 with engine_forkchoiceUpdatedV1. 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:

  • parent_hash
  • timestamp
  • randao
  • fee_recipient
  • transactions_list (first length, then equality of each of the encoded transactions, including deposits)

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_forkchoiceUpdatedV1 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_getPayload to retrieve the payload, by the payload-ID in the result of the previous step.
  • engine_newPayload to import the new payload into the execution engine.
  • engine_forkchoiceUpdatedV1 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:

  • engine_newPayloadV1: process the payload. It does not become canonical yet.
  • engine_forkchoiceUpdatedV1: 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 L2 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 PayloadAttributesV1 object, which includes additional transactions and noTxPool fields.

This process happens during the payloads-attributes queue ran by a full node or validator 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 PayloadAttributesV1 object must include the following transactions:

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.

Building Individual Payload Attributes

After deriving the transactions list, the rollup node constructs a PayloadAttributesV1 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 zero address because the transaction fee is distributed to the fee vaults. 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.