feat: consider RLE blocks in compatibility check starting from V9 (#67)

* fix: consider RLE blocks in zstd compatibility check (#64)

* fix compability

* fmt

* override behavior

* Apply suggestions from code review

Co-authored-by: Copilot <[email protected]>

---------

Co-authored-by: Ho <[email protected]>
Co-authored-by: georgehao <[email protected]>
Co-authored-by: Copilot <[email protected]>

Author: 4 people

encoding/codecv9.go

@@ -1,5 +1,31 @@
 package encoding
 
+import (
+	"crypto/sha256"
+	"encoding/hex"
+	"errors"
+	"fmt"
+
+	"github.com/scroll-tech/go-ethereum/common"
+	"github.com/scroll-tech/go-ethereum/crypto"
+	"github.com/scroll-tech/go-ethereum/crypto/kzg4844"
+	"github.com/scroll-tech/go-ethereum/log"
+)
+
+// DACodecV9 updates the implementation of the base function checkCompressedDataCompatibility
+// to use the V9 compatibility check (checkCompressedDataCompatibilityV9) instead of the previous V7 version.
+//
+// As per Go's rules for shadowing methods with struct embedding, we need to override
+// all methods that (transitively) call checkCompressedDataCompatibility:
+// - checkCompressedDataCompatibility (updated to use V9)
+// - constructBlob (calls checkCompressedDataCompatibility)
+// - NewDABatch (calls constructBlob)
+// - CheckChunkCompressedDataCompatibility (calls CheckBatchCompressedDataCompatibility)
+// - CheckBatchCompressedDataCompatibility (calls checkCompressedDataCompatibility)
+// - estimateL1CommitBatchSizeAndBlobSize (calls checkCompressedDataCompatibility)
+// - EstimateChunkL1CommitBatchSizeAndBlobSize (calls estimateL1CommitBatchSizeAndBlobSize)
+// - EstimateBatchL1CommitBatchSizeAndBlobSize (calls estimateL1CommitBatchSizeAndBlobSize)
+
 type DACodecV9 struct {
 	DACodecV8
 }
@@ -12,3 +38,184 @@ func NewDACodecV9() *DACodecV9 {
 		},
 	}
 }
+
+// checkCompressedDataCompatibility checks the compressed data compatibility for a batch.
+// It constructs a blob payload, compresses the data, and checks the compressed data compatibility.
+// flag checkLength indicates whether to check the length of the compressed data against the original data.
+// If checkLength is true, this function returns if compression is needed based on the compressed data's length, which is used when doing batch bytes encoding.
+// If checkLength is false, this function returns the result of the compatibility check, which is used when determining the chunk and batch contents.
+func (d *DACodecV9) checkCompressedDataCompatibility(payloadBytes []byte, checkLength bool) ([]byte, bool, error) {
+	compressedPayloadBytes, err := d.CompressScrollBatchBytes(payloadBytes)
+	if err != nil {
+		return nil, false, fmt.Errorf("failed to compress blob payload: %w", err)
+	}
+
+	if err = checkCompressedDataCompatibilityV9(compressedPayloadBytes); err != nil {
+		log.Warn("Compressed data compatibility check failed", "err", err, "payloadBytes", hex.EncodeToString(payloadBytes), "compressedPayloadBytes", hex.EncodeToString(compressedPayloadBytes))
+		return nil, false, nil
+	}
+
+	// check if compressed data is bigger or equal to the original data -> no need to compress
+	if checkLength && len(compressedPayloadBytes) >= len(payloadBytes) {
+		log.Warn("Compressed data is bigger or equal to the original data", "payloadBytes", hex.EncodeToString(payloadBytes), "compressedPayloadBytes", hex.EncodeToString(compressedPayloadBytes))
+		return nil, false, nil
+	}
+
+	return compressedPayloadBytes, true, nil
+}
+
+// NewDABatch creates a DABatch including blob from the provided Batch.
+func (d *DACodecV9) NewDABatch(batch *Batch) (DABatch, error) {
+	if len(batch.Blocks) == 0 {
+		return nil, errors.New("batch must contain at least one block")
+	}
+
+	if err := checkBlocksBatchVSChunksConsistency(batch); err != nil {
+		return nil, fmt.Errorf("failed to check blocks batch vs chunks consistency: %w", err)
+	}
+
+	blob, blobVersionedHash, blobBytes, challengeDigest, err := d.constructBlob(batch)
+	if err != nil {
+		return nil, fmt.Errorf("failed to construct blob: %w", err)
+	}
+
+	daBatch, err := newDABatchV7(d.Version(), batch.Index, blobVersionedHash, batch.ParentBatchHash, blob, blobBytes, challengeDigest)
+	if err != nil {
+		return nil, fmt.Errorf("failed to construct DABatch: %w", err)
+	}
+
+	return daBatch, nil
+}
+
+func (d *DACodecV9) constructBlob(batch *Batch) (*kzg4844.Blob, common.Hash, []byte, common.Hash, error) {
+	blobBytes := make([]byte, blobEnvelopeV7OffsetPayload)
+
+	payloadBytes, err := d.constructBlobPayload(batch)
+	if err != nil {
+		return nil, common.Hash{}, nil, common.Hash{}, fmt.Errorf("failed to construct blob payload: %w", err)
+	}
+
+	compressedPayloadBytes, enableCompression, err := d.checkCompressedDataCompatibility(payloadBytes, true /* checkLength */)
+	if err != nil {
+		return nil, common.Hash{}, nil, common.Hash{}, fmt.Errorf("failed to check batch compressed data compatibility: %w", err)
+	}
+
+	isCompressedFlag := uint8(0x0)
+	if enableCompression {
+		isCompressedFlag = 0x1
+		payloadBytes = compressedPayloadBytes
+	}
+
+	sizeSlice := encodeSize3Bytes(uint32(len(payloadBytes)))
+
+	blobBytes[blobEnvelopeV7OffsetVersion] = uint8(d.Version())
+	copy(blobBytes[blobEnvelopeV7OffsetByteSize:blobEnvelopeV7OffsetCompressedFlag], sizeSlice)
+	blobBytes[blobEnvelopeV7OffsetCompressedFlag] = isCompressedFlag
+	blobBytes = append(blobBytes, payloadBytes...)
+
+	if len(blobBytes) > maxEffectiveBlobBytes {
+		log.Error("ConstructBlob: Blob payload exceeds maximum size", "size", len(blobBytes), "blobBytes", hex.EncodeToString(blobBytes))
+		return nil, common.Hash{}, nil, common.Hash{}, fmt.Errorf("blob exceeds maximum size: got %d, allowed %d", len(blobBytes), maxEffectiveBlobBytes)
+	}
+
+	// convert raw data to BLSFieldElements
+	blob, err := makeBlobCanonical(blobBytes)
+	if err != nil {
+		return nil, common.Hash{}, nil, common.Hash{}, fmt.Errorf("failed to convert blobBytes to canonical form: %w", err)
+	}
+
+	// compute blob versioned hash
+	c, err := kzg4844.BlobToCommitment(blob)
+	if err != nil {
+		return nil, common.Hash{}, nil, common.Hash{}, fmt.Errorf("failed to create blob commitment: %w", err)
+	}
+	blobVersionedHash := kzg4844.CalcBlobHashV1(sha256.New(), &c)
+
+	// compute challenge digest for codecv7, different from previous versions,
+	// the blob bytes are padded to the max effective blob size, which is 131072 / 32 * 31 due to the blob encoding
+	paddedBlobBytes := make([]byte, maxEffectiveBlobBytes)
+	copy(paddedBlobBytes, blobBytes)
+
+	challengeDigest := crypto.Keccak256Hash(crypto.Keccak256(paddedBlobBytes), blobVersionedHash[:])
+
+	return blob, blobVersionedHash, blobBytes, challengeDigest, nil
+}
+
+// CheckChunkCompressedDataCompatibility checks the compressed data compatibility for a batch built from a single chunk.
+func (d *DACodecV9) CheckChunkCompressedDataCompatibility(c *Chunk) (bool, error) {
+	// filling the needed fields for the batch used in the check
+	b := &Batch{
+		Chunks:                 []*Chunk{c},
+		PrevL1MessageQueueHash: c.PrevL1MessageQueueHash,
+		PostL1MessageQueueHash: c.PostL1MessageQueueHash,
+		Blocks:                 c.Blocks,
+	}
+
+	return d.CheckBatchCompressedDataCompatibility(b)
+}
+
+// CheckBatchCompressedDataCompatibility checks the compressed data compatibility for a batch.
+func (d *DACodecV9) CheckBatchCompressedDataCompatibility(b *Batch) (bool, error) {
+	if len(b.Blocks) == 0 {
+		return false, errors.New("batch must contain at least one block")
+	}
+
+	if err := checkBlocksBatchVSChunksConsistency(b); err != nil {
+		return false, fmt.Errorf("failed to check blocks batch vs chunks consistency: %w", err)
+	}
+
+	payloadBytes, err := d.constructBlobPayload(b)
+	if err != nil {
+		return false, fmt.Errorf("failed to construct blob payload: %w", err)
+	}
+
+	// This check is only used for sanity checks. If the check fails, it means that the compression did not work as expected.
+	// rollup-relayer will try popping the last chunk of the batch (or last block of the chunk when in proposing chunks) and try again to see if it works as expected.
+	// Since length check is used for DA and proving efficiency, it does not need to be checked here.
+	_, compatible, err := d.checkCompressedDataCompatibility(payloadBytes, false /* checkLength */)
+	if err != nil {
+		return false, fmt.Errorf("failed to check batch compressed data compatibility: %w", err)
+	}
+
+	return compatible, nil
+}
+
+func (d *DACodecV9) estimateL1CommitBatchSizeAndBlobSize(batch *Batch) (uint64, uint64, error) {
+	if len(batch.Blocks) == 0 {
+		return 0, 0, errors.New("batch must contain at least one block")
+	}
+
+	blobBytes := make([]byte, blobEnvelopeV7OffsetPayload)
+
+	payloadBytes, err := d.constructBlobPayload(batch)
+	if err != nil {
+		return 0, 0, fmt.Errorf("failed to construct blob payload: %w", err)
+	}
+
+	compressedPayloadBytes, enableCompression, err := d.checkCompressedDataCompatibility(payloadBytes, true /* checkLength */)
+	if err != nil {
+		return 0, 0, fmt.Errorf("failed to check batch compressed data compatibility: %w", err)
+	}
+
+	if enableCompression {
+		blobBytes = append(blobBytes, compressedPayloadBytes...)
+	} else {
+		blobBytes = append(blobBytes, payloadBytes...)
+	}
+
+	return blobEnvelopeV7OffsetPayload + uint64(len(payloadBytes)), calculatePaddedBlobSize(uint64(len(blobBytes))), nil
+}
+
+// EstimateChunkL1CommitBatchSizeAndBlobSize estimates the L1 commit batch size and blob size for a single chunk.
+func (d *DACodecV9) EstimateChunkL1CommitBatchSizeAndBlobSize(chunk *Chunk) (uint64, uint64, error) {
+	return d.estimateL1CommitBatchSizeAndBlobSize(&Batch{
+		Blocks:                 chunk.Blocks,
+		PrevL1MessageQueueHash: chunk.PrevL1MessageQueueHash,
+		PostL1MessageQueueHash: chunk.PostL1MessageQueueHash,
+	})
+}
+
+// EstimateBatchL1CommitBatchSizeAndBlobSize estimates the L1 commit batch size and blob size for a batch.
+func (d *DACodecV9) EstimateBatchL1CommitBatchSizeAndBlobSize(batch *Batch) (uint64, uint64, error) {
+	return d.estimateL1CommitBatchSizeAndBlobSize(batch)
+}

encoding/da.go

@@ -509,6 +509,61 @@ func checkCompressedDataCompatibilityV7(data []byte) error {
 	return nil
 }
 
+// Sanity check if the compressed data (v9) is compatible with our circuit.
+// If we conclude that the data could not be decompressed, then we will
+// commit it uncompressed instead.
+func checkCompressedDataCompatibilityV9(data []byte) error {
+	if len(data) < 16 {
+		return fmt.Errorf("too small size (0x%x), what is it?", data)
+	}
+
+	fheader := data[0]
+	// it is not the encoding type we expected in our zstd header
+	if fheader&63 != 32 {
+		return fmt.Errorf("unexpected header type (%x)", fheader)
+	}
+
+	// skip content size
+	switch fheader >> 6 {
+	case 0:
+		data = data[2:]
+	case 1:
+		data = data[3:]
+	case 2:
+		data = data[5:]
+	case 3:
+		data = data[9:]
+	default:
+		panic("impossible")
+	}
+
+	isLast := false
+	// scan each block until done
+	for len(data) > 3 && !isLast {
+		isLast = (data[0] & 1) == 1
+		blkType := (data[0] >> 1) & 3
+		var blkSize uint
+		if blkType == 1 { // RLE Block
+			blkSize = 1
+		} else {
+			if blkType == 3 {
+				return fmt.Errorf("encounter reserved block type at %v", data)
+			}
+			blkSize = (uint(data[2])*65536 + uint(data[1])*256 + uint(data[0])) >> 3
+		}
+		if len(data) < 3+int(blkSize) {
+			return fmt.Errorf("wrong data len {%d}, expect min {%d}", len(data), 3+blkSize)
+		}
+		data = data[3+blkSize:]
+	}
+
+	if !isLast {
+		return fmt.Errorf("unexpected end before last block")
+	}
+
+	return nil
+}
+
 // makeBlobCanonical converts the raw blob data into the canonical blob representation of 4096 BLSFieldElements.
 // The canonical blob representation is a 32-byte array where every 31 bytes are prepended with 1 zero byte.
 // The kzg4844.Blob is a 4096-byte array, thus 0s are padded to the end of the array.