beacon/engine, rpc: optimize JSON encoding for large blob payloads (#33969)

Adds a fast path for ExecutionPayloadEnvelope and BlobAndProofListV*
that bypasses encoding/json's reflection and re-validation, which are
expensive for large payloads with many blobs. Also hand-rolls the
jsonrpcMessage wire encoding in the RPC codec to avoid a second
re-validation pass when writing responses to the connection.

Resolves #33814

---------

Co-authored-by: Marius van der Wijden <m.vanderwijden@live.de>
Co-authored-by: Felix Lange <fjl@twurst.com>

beacon/engine/bapl_encode.go

@@ -0,0 +1,69 @@
+// Copyright 2026 The go-ethereum Authors
+// This file is part of the go-ethereum library.
+//
+// The go-ethereum library is free software: you can redistribute it and/or modify
+// it under the terms of the GNU Lesser General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+//
+// The go-ethereum library is distributed in the hope that it will be useful,
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+// GNU Lesser General Public License for more details.
+//
+// You should have received a copy of the GNU Lesser General Public License
+// along with the go-ethereum library. If not, see <http://www.gnu.org/licenses/>.
+
+package engine
+
+import (
+	"github.com/fjl/jsonw"
+)
+
+// MarshalJSON implements json.Marshaler.
+func (list BlobAndProofListV1) MarshalJSON() ([]byte, error) {
+	var b jsonw.Buffer
+	b.Array(func() {
+		for _, item := range list {
+			marshalBlobAndProofV1(&b, item)
+		}
+	})
+	return b.Output(), nil
+}
+
+func marshalBlobAndProofV1(b *jsonw.Buffer, item *BlobAndProofV1) {
+	if item == nil {
+		b.Null()
+	} else {
+		b.Object(func() {
+			b.Key("blob")
+			b.HexBytes(item.Blob)
+			b.Key("proof")
+			b.HexBytes(item.Proof)
+		})
+	}
+}
+
+// MarshalJSON implements json.Marshaler.
+func (list BlobAndProofListV2) MarshalJSON() ([]byte, error) {
+	var b jsonw.Buffer
+	b.Array(func() {
+		for _, item := range list {
+			marshalBlobAndProofV2(&b, item)
+		}
+	})
+	return b.Output(), nil
+}
+
+func marshalBlobAndProofV2(b *jsonw.Buffer, item *BlobAndProofV2) {
+	if item == nil {
+		b.Null()
+	} else {
+		b.Object(func() {
+			b.Key("blob")
+			b.HexBytes(item.Blob)
+			b.Key("proofs")
+			appendHexBytesArray(b, item.CellProofs)
+		})
+	}
+}

beacon/engine/epe_decode.go

@@ -12,31 +12,6 @@ import (
 
 var _ = (*executionPayloadEnvelopeMarshaling)(nil)
 
-// MarshalJSON marshals as JSON.
-func (e ExecutionPayloadEnvelope) MarshalJSON() ([]byte, error) {
-	type ExecutionPayloadEnvelope struct {
-		ExecutionPayload *ExecutableData `json:"executionPayload"  gencodec:"required"`
-		BlockValue       *hexutil.Big    `json:"blockValue"  gencodec:"required"`
-		BlobsBundle      *BlobsBundle    `json:"blobsBundle"`
-		Requests         []hexutil.Bytes `json:"executionRequests"`
-		Override         bool            `json:"shouldOverrideBuilder"`
-		Witness          *hexutil.Bytes  `json:"witness,omitempty"`
-	}
-	var enc ExecutionPayloadEnvelope
-	enc.ExecutionPayload = e.ExecutionPayload
-	enc.BlockValue = (*hexutil.Big)(e.BlockValue)
-	enc.BlobsBundle = e.BlobsBundle
-	if e.Requests != nil {
-		enc.Requests = make([]hexutil.Bytes, len(e.Requests))
-		for k, v := range e.Requests {
-			enc.Requests[k] = v
-		}
-	}
-	enc.Override = e.Override
-	enc.Witness = e.Witness
-	return json.Marshal(&enc)
-}
-
 // UnmarshalJSON unmarshals from JSON.
 func (e *ExecutionPayloadEnvelope) UnmarshalJSON(input []byte) error {
 	type ExecutionPayloadEnvelope struct {

beacon/engine/epe_encode.go

@@ -0,0 +1,98 @@
+// Copyright 2026 The go-ethereum Authors
+// This file is part of the go-ethereum library.
+//
+// The go-ethereum library is free software: you can redistribute it and/or modify
+// it under the terms of the GNU Lesser General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+//
+// The go-ethereum library is distributed in the hope that it will be useful,
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+// GNU Lesser General Public License for more details.
+//
+// You should have received a copy of the GNU Lesser General Public License
+// along with the go-ethereum library. If not, see <http://www.gnu.org/licenses/>.
+
+package engine
+
+import (
+	"encoding/json"
+	"errors"
+
+	"github.com/fjl/jsonw"
+)
+
+// marshalBlobsBundle writes BlobsBundle as JSON and appends it to buf.
+func marshalBlobsBundle(b *jsonw.Buffer, bundle *BlobsBundle) {
+	if bundle == nil {
+		b.Null()
+		return
+	}
+	b.Object(func() {
+		b.Key("commitments")
+		appendHexBytesArray(b, bundle.Commitments)
+		b.Key("proofs")
+		appendHexBytesArray(b, bundle.Proofs)
+		b.Key("blobs")
+		appendHexBytesArray(b, bundle.Blobs)
+	})
+}
+
+// MarshalJSON implements json.Marshaler.
+func (e ExecutionPayloadEnvelope) MarshalJSON() ([]byte, error) {
+	if e.ExecutionPayload == nil {
+		return nil, errors.New("missing required field 'executionPayload' for ExecutionPayloadEnvelope")
+	}
+
+	// Pre-marshal the execution payload using its gencodec MarshalJSON.
+	payload, err := e.ExecutionPayload.MarshalJSON()
+	if err != nil {
+		return nil, err
+	}
+	// Pre-marshal the witness.
+	var witness []byte
+	if e.Witness != nil {
+		witness, err = json.Marshal(e.Witness)
+		if err != nil {
+			return nil, err
+		}
+	}
+
+	// Write the execution payload to the buffer
+	var b jsonw.Buffer
+	b.Object(func() {
+		b.Key("executionPayload")
+		b.RawValue(payload)
+
+		b.Key("blockValue")
+		b.HexBigInt(e.BlockValue)
+
+		b.Key("blobsBundle")
+		marshalBlobsBundle(&b, e.BlobsBundle)
+
+		b.Key("executionRequests")
+		if e.Requests == nil {
+			b.Null()
+		} else {
+			appendHexBytesArray(&b, e.Requests)
+		}
+
+		b.Key("shouldOverrideBuilder")
+		b.Bool(e.Override)
+
+		if e.Witness != nil {
+			b.Key("witness")
+			b.RawValue(witness)
+		}
+	})
+	return b.Output(), nil
+}
+
+func appendHexBytesArray[T ~[]byte](b *jsonw.Buffer, slice []T) {
+	b.Array(func() {
+		for _, elem := range slice {
+			b.HexBytes(elem)
+		}
+	})
+}

beacon/engine/epe_test.go

@@ -0,0 +1,128 @@
+// Copyright 2026 The go-ethereum Authors
+// This file is part of the go-ethereum library.
+//
+// The go-ethereum library is free software: you can redistribute it and/or modify
+// it under the terms of the GNU Lesser General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+//
+// The go-ethereum library is distributed in the hope that it will be useful,
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+// GNU Lesser General Public License for more details.
+//
+// You should have received a copy of the GNU Lesser General Public License
+// along with the go-ethereum library. If not, see <http://www.gnu.org/licenses/>.
+
+package engine
+
+import (
+	"bytes"
+	"encoding/json"
+	"math/big"
+	"reflect"
+	"testing"
+
+	"github.com/ethereum/go-ethereum/common"
+	"github.com/ethereum/go-ethereum/common/hexutil"
+)
+
+func makeTestPayload() *ExecutableData {
+	return &ExecutableData{
+		ParentHash:    common.HexToHash("0x01"),
+		FeeRecipient:  common.HexToAddress("0x02"),
+		StateRoot:     common.HexToHash("0x03"),
+		ReceiptsRoot:  common.HexToHash("0x04"),
+		LogsBloom:     make([]byte, 256),
+		Random:        common.HexToHash("0x05"),
+		Number:        100,
+		GasLimit:      1000000,
+		GasUsed:       500000,
+		Timestamp:     1234567890,
+		ExtraData:     []byte("extra"),
+		BaseFeePerGas: big.NewInt(7),
+		BlockHash:     common.HexToHash("0x08"),
+		Transactions:  [][]byte{{0xaa, 0xbb}},
+	}
+}
+
+func TestMarshalJSONRoundtrip(t *testing.T) {
+	witness := hexutil.Bytes{0xde, 0xad}
+	original := ExecutionPayloadEnvelope{
+		ExecutionPayload: makeTestPayload(),
+		BlockValue:       big.NewInt(12345),
+		BlobsBundle: &BlobsBundle{
+			Commitments: []hexutil.Bytes{{0x01, 0x02}},
+			Proofs:      []hexutil.Bytes{{0x03, 0x04}},
+			Blobs:       []hexutil.Bytes{{0x05, 0x06}},
+		},
+		Requests: [][]byte{{0xaa}, {0xbb, 0xcc}},
+		Override: true,
+		Witness:  &witness,
+	}
+
+	data, err := original.MarshalJSON()
+	if err != nil {
+		t.Fatalf("MarshalJSON error: %v", err)
+	}
+
+	var decoded ExecutionPayloadEnvelope
+	if err := json.Unmarshal(data, &decoded); err != nil {
+		t.Fatalf("UnmarshalJSON error: %v", err)
+	}
+
+	if decoded.ExecutionPayload.Number != original.ExecutionPayload.Number {
+		t.Error("ExecutionPayload.Number mismatch")
+	}
+	if decoded.BlockValue.Cmp(original.BlockValue) != 0 {
+		t.Errorf("BlockValue mismatch: got %v, want %v", decoded.BlockValue, original.BlockValue)
+	}
+	if len(decoded.BlobsBundle.Blobs) != len(original.BlobsBundle.Blobs) {
+		t.Error("BlobsBundle.Blobs length mismatch")
+	}
+	if len(decoded.Requests) != len(original.Requests) {
+		t.Error("Requests length mismatch")
+	}
+	if decoded.Override != original.Override {
+		t.Error("Override mismatch")
+	}
+	if !bytes.Equal(*decoded.Witness, *original.Witness) {
+		t.Error("Witness mismatch")
+	}
+}
+
+func TestMarshalJSONNilPayload(t *testing.T) {
+	env := ExecutionPayloadEnvelope{
+		ExecutionPayload: nil,
+		BlockValue:       big.NewInt(1),
+	}
+	_, err := env.MarshalJSON()
+	if err == nil {
+		t.Fatal("expected error for nil ExecutionPayload")
+	}
+}
+
+// TestExecutionPayloadEnvelopeFieldCoverage guards against structural drift.
+// If a field is added to or removed from ExecutionPayloadEnvelope, this test
+// fails, reminding the developer to update MarshalJSON in marshal_epe.go.
+func TestExecutionPayloadEnvelopeFieldCoverage(t *testing.T) {
+	expected := []string{
+		"ExecutionPayload",
+		"BlockValue",
+		"BlobsBundle",
+		"Requests",
+		"Override",
+		"Witness",
+	}
+	typ := reflect.TypeOf(ExecutionPayloadEnvelope{})
+	if typ.NumField() != len(expected) {
+		t.Fatalf("ExecutionPayloadEnvelope has %d fields, expected %d — update MarshalJSON in marshal_epe.go",
+			typ.NumField(), len(expected))
+	}
+	for i, name := range expected {
+		if typ.Field(i).Name != name {
+			t.Errorf("field %d: got %q, want %q — update MarshalJSON in marshal_epe.go",
+				i, typ.Field(i).Name, name)
+		}
+	}
+}

beacon/engine/types.go

@@ -60,7 +60,7 @@ var (
 	PayloadV4 PayloadVersion = 0x4
 )
 
-//go:generate go run github.com/fjl/gencodec -type PayloadAttributes -field-override payloadAttributesMarshaling -out gen_blockparams.go
+//go:generate go run github.com/fjl/gencodec -type PayloadAttributes -field-override payloadAttributesMarshaling -out pa_codec.go
 
 // PayloadAttributes describes the environment context in which a block should
 // be built.
@@ -79,7 +79,7 @@ type payloadAttributesMarshaling struct {
 	SlotNumber *hexutil.Uint64
 }
 
-//go:generate go run github.com/fjl/gencodec -type ExecutableData -field-override executableDataMarshaling -out gen_ed.go
+//go:generate go run github.com/fjl/gencodec -type ExecutableData -field-override executableDataMarshaling -out ed_codec.go
 
 // ExecutableData is the data necessary to execute an EL payload.
 type ExecutableData struct {
@@ -127,7 +127,7 @@ type StatelessPayloadStatusV1 struct {
 	ValidationError *string     `json:"validationError"`
 }
 
-//go:generate go run github.com/fjl/gencodec -type ExecutionPayloadEnvelope -field-override executionPayloadEnvelopeMarshaling -out gen_epe.go
+//go:generate go run github.com/fjl/gencodec -enc=false -type ExecutionPayloadEnvelope -field-override executionPayloadEnvelopeMarshaling -out epe_decode.go
 
 type ExecutionPayloadEnvelope struct {
 	ExecutionPayload *ExecutableData `json:"executionPayload"  gencodec:"required"`
@@ -138,6 +138,12 @@ type ExecutionPayloadEnvelope struct {
 	Witness          *hexutil.Bytes  `json:"witness,omitempty"`
 }
 
+// JSON type overrides for ExecutionPayloadEnvelope.
+type executionPayloadEnvelopeMarshaling struct {
+	BlockValue *hexutil.Big
+	Requests   []hexutil.Bytes
+}
+
 // BlobsBundle includes the marshalled sidecar data. Note this structure is
 // shared by BlobsBundleV1 and BlobsBundleV2 for the sake of simplicity.
 //
@@ -154,16 +160,18 @@ type BlobAndProofV1 struct {
 	Proof hexutil.Bytes `json:"proof"`
 }
 
+// BlobAndProofListV1 is a list of BlobAndProofV1 with a hand-rolled JSON marshaler
+// that avoids the overhead of encoding/json for large blob payloads.
+type BlobAndProofListV1 []*BlobAndProofV1
+
 type BlobAndProofV2 struct {
 	Blob       hexutil.Bytes   `json:"blob"`
 	CellProofs []hexutil.Bytes `json:"proofs"` // proofs MUST contain exactly CELLS_PER_EXT_BLOB cell proofs.
 }
 
-// JSON type overrides for ExecutionPayloadEnvelope.
-type executionPayloadEnvelopeMarshaling struct {
-	BlockValue *hexutil.Big
-	Requests   []hexutil.Bytes
-}
+// BlobAndProofListV2 is a list of BlobAndProofV2 with a hand-rolled JSON marshaler
+// that avoids the overhead of encoding/json for large blob payloads.
+type BlobAndProofListV2 []*BlobAndProofV2
 
 type PayloadStatusV1 struct {
 	Status          string         `json:"status"`

eth/catalyst/api.go

@@ -552,7 +552,7 @@ func (api *ConsensusAPI) getPayload(payloadID engine.PayloadID, full bool, versi
 //
 // Client software MAY return an array of all null entries if syncing or otherwise
 // unable to serve blob pool data.
-func (api *ConsensusAPI) GetBlobsV1(hashes []common.Hash) ([]*engine.BlobAndProofV1, error) {
+func (api *ConsensusAPI) GetBlobsV1(hashes []common.Hash) (engine.BlobAndProofListV1, error) {
 	// Reject the request if Osaka has been activated.
 	// follow https://github.com/ethereum/execution-apis/blob/main/src/engine/osaka.md#cancun-api
 	head := api.eth.BlockChain().CurrentHeader()
@@ -566,7 +566,7 @@ func (api *ConsensusAPI) GetBlobsV1(hashes []common.Hash) ([]*engine.BlobAndProo
 	if err != nil {
 		return nil, engine.InvalidParams.With(err)
 	}
-	res := make([]*engine.BlobAndProofV1, len(hashes))
+	res := make(engine.BlobAndProofListV1, len(hashes))
 	for i := 0; i < len(blobs); i++ {
 		// Skip the non-existing blob
 		if blobs[i] == nil {
@@ -605,7 +605,7 @@ func (api *ConsensusAPI) GetBlobsV1(hashes []common.Hash) ([]*engine.BlobAndProo
 //
 // Client software MUST return null if syncing or otherwise unable to serve
 // blob pool data.
-func (api *ConsensusAPI) GetBlobsV2(hashes []common.Hash) ([]*engine.BlobAndProofV2, error) {
+func (api *ConsensusAPI) GetBlobsV2(hashes []common.Hash) (engine.BlobAndProofListV2, error) {
 	head := api.eth.BlockChain().CurrentHeader()
 	if api.config().LatestFork(head.Time) < forks.Osaka {
 		return nil, nil
@@ -616,7 +616,7 @@ func (api *ConsensusAPI) GetBlobsV2(hashes []common.Hash) ([]*engine.BlobAndProo
 // GetBlobsV3 returns a set of blobs from the transaction pool. Same as
 // GetBlobsV2, except will return partial responses in case there is a missing
 // blob.
-func (api *ConsensusAPI) GetBlobsV3(hashes []common.Hash) ([]*engine.BlobAndProofV2, error) {
+func (api *ConsensusAPI) GetBlobsV3(hashes []common.Hash) (engine.BlobAndProofListV2, error) {
 	head := api.eth.BlockChain().CurrentHeader()
 	if api.config().LatestFork(head.Time) < forks.Osaka {
 		return nil, nil
@@ -626,7 +626,7 @@ func (api *ConsensusAPI) GetBlobsV3(hashes []common.Hash) ([]*engine.BlobAndProo
 
 // getBlobs returns all available blobs. In v2, partial responses are not allowed,
 // while v3 supports partial responses.
-func (api *ConsensusAPI) getBlobs(hashes []common.Hash, v2 bool) ([]*engine.BlobAndProofV2, error) {
+func (api *ConsensusAPI) getBlobs(hashes []common.Hash, v2 bool) (engine.BlobAndProofListV2, error) {
 	if len(hashes) > 128 {
 		return nil, engine.TooLargeRequest.With(fmt.Errorf("requested blob count too large: %v", len(hashes)))
 	}
@@ -647,7 +647,7 @@ func (api *ConsensusAPI) getBlobs(hashes []common.Hash, v2 bool) ([]*engine.Blob
 	}
 	// Validate the blobs from the pool and assemble the response
 	filled := 0
-	res := make([]*engine.BlobAndProofV2, len(hashes))
+	res := make(engine.BlobAndProofListV2, len(hashes))
 	for i := range blobs {
 		// The blob has been evicted since the last AvailableBlobs call.
 		// Return null if partial response is not allowed.

eth/catalyst/api_benchmark_test.go

@@ -0,0 +1,739 @@
+// Copyright 2026 The go-ethereum Authors
+// This file is part of the go-ethereum library.
+//
+// The go-ethereum library is free software: you can redistribute it and/or modify
+// it under the terms of the GNU Lesser General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+//
+// The go-ethereum library is distributed in the hope that it will be useful,
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+// GNU Lesser General Public License for more details.
+//
+// You should have received a copy of the GNU Lesser General Public License
+// along with the go-ethereum library. If not, see <http://www.gnu.org/licenses/>.
+package catalyst
+
+import (
+	"context"
+	"crypto/ecdsa"
+	"crypto/sha256"
+	"encoding/json"
+	"fmt"
+	"io"
+	"math/big"
+	"strings"
+	"testing"
+	"time"
+
+	"github.com/ethereum/go-ethereum/beacon/engine"
+	"github.com/ethereum/go-ethereum/common"
+	"github.com/ethereum/go-ethereum/core"
+	"github.com/ethereum/go-ethereum/core/types"
+	"github.com/ethereum/go-ethereum/crypto"
+	"github.com/ethereum/go-ethereum/crypto/kzg4844"
+	"github.com/ethereum/go-ethereum/eth"
+	"github.com/ethereum/go-ethereum/node"
+	"github.com/ethereum/go-ethereum/params"
+	"github.com/ethereum/go-ethereum/rlp"
+	"github.com/ethereum/go-ethereum/rpc"
+	"github.com/holiman/uint256"
+)
+
+// encodingType specifies which encoding to use in benchmarks
+type encodingType int
+
+const (
+	encNone encodingType = iota
+	encJSON
+	encJSONCustom
+	encRLP
+)
+
+func (e encodingType) String() string {
+	switch e {
+	case encNone:
+		return "none"
+	case encJSON:
+		return "json"
+	case encJSONCustom:
+		return "json_custom"
+	case encRLP:
+		return "rlp"
+	default:
+		return "unknown"
+	}
+}
+
+var encodingTypes = []encodingType{encNone, encJSON, encJSONCustom, encRLP}
+
+// benchEncode encodes the value using the specified encoding type.
+// It fails the benchmark if encoding fails.
+func benchEncode(b *testing.B, enc encodingType, v any) {
+	var err error
+	switch enc {
+	case encJSON:
+		_, err = json.Marshal(v)
+		if err != nil {
+			b.Fatalf("JSON marshal failed: %v", err)
+		}
+	case encJSONCustom:
+		if m, ok := v.(json.Marshaler); ok {
+			_, err = m.MarshalJSON()
+		} else {
+			_, err = json.Marshal(v)
+		}
+		if err != nil {
+			b.Fatalf("JSON MarshalJSON failed: %v", err)
+		}
+	case encRLP:
+		_, err = rlp.EncodeToBytes(v)
+		if err != nil {
+			b.Fatalf("RLP encode failed: %v", err)
+		}
+	}
+}
+
+// benchmarkBlobCounts defines the blob counts for benchmarks
+var benchmarkBlobCounts = []int{21, 72}
+
+// maxBenchmarkBlobs is the maximum number of blobs we need for benchmarks
+var maxBenchmarkBlobs = benchmarkBlobCounts[len(benchmarkBlobCounts)-1]
+
+var (
+	// Pre-computed blobs for benchmarks
+	benchBlobs          []*kzg4844.Blob
+	benchBlobCommits    []kzg4844.Commitment
+	benchBlobProofs     []kzg4844.Proof
+	benchBlobCellProofs [][]kzg4844.Proof
+	benchBlobVHashes    []common.Hash
+)
+
+func init() {
+	// Pre-compute blobs for benchmarks
+	for i := 0; i < maxBenchmarkBlobs; i++ {
+		blob := &kzg4844.Blob{byte(i), byte(i >> 8)}
+		benchBlobs = append(benchBlobs, blob)
+
+		commit, _ := kzg4844.BlobToCommitment(blob)
+		benchBlobCommits = append(benchBlobCommits, commit)
+
+		proof, _ := kzg4844.ComputeBlobProof(blob, commit)
+		benchBlobProofs = append(benchBlobProofs, proof)
+
+		cellProofs, _ := kzg4844.ComputeCellProofs(blob)
+		benchBlobCellProofs = append(benchBlobCellProofs, cellProofs)
+
+		vhash := kzg4844.CalcBlobHashV1(sha256.New(), &commit)
+		benchBlobVHashes = append(benchBlobVHashes, vhash)
+	}
+}
+
+// benchFork specifies which fork to use in benchmark environments
+type benchFork int
+
+const (
+	forkCancun benchFork = iota
+	forkPrague
+	forkOsaka
+)
+
+// benchmarkBlobEnv holds the environment for blob benchmarks
+type benchmarkBlobEnv struct {
+	node      *node.Node
+	eth       *eth.Ethereum
+	api       *ConsensusAPI
+	config    *params.ChainConfig
+	keys      []*ecdsa.PrivateKey
+	vhashes   []common.Hash
+	version   byte
+	blobCount int
+	nonces    []uint64 // current nonce for each key
+}
+
+// makeBenchBlobTx creates a blob transaction with the specified number of blobs.
+// blobOffset indicates which pre-computed blobs to use.
+func makeBenchBlobTx(chainConfig *params.ChainConfig, nonce uint64, blobCount int, blobOffset int, key *ecdsa.PrivateKey, version byte) *types.Transaction {
+	var (
+		blobs       []kzg4844.Blob
+		blobHashes  []common.Hash
+		commitments []kzg4844.Commitment
+		proofs      []kzg4844.Proof
+	)
+	for i := 0; i < blobCount; i++ {
+		idx := blobOffset + i
+		blobs = append(blobs, *benchBlobs[idx])
+		commitments = append(commitments, benchBlobCommits[idx])
+		if version == types.BlobSidecarVersion0 {
+			proofs = append(proofs, benchBlobProofs[idx])
+		} else {
+			proofs = append(proofs, benchBlobCellProofs[idx]...)
+		}
+		blobHashes = append(blobHashes, benchBlobVHashes[idx])
+	}
+	blobtx := &types.BlobTx{
+		ChainID:    uint256.MustFromBig(chainConfig.ChainID),
+		Nonce:      nonce,
+		GasTipCap:  uint256.NewInt(params.GWei),
+		GasFeeCap:  uint256.NewInt(10 * params.GWei),
+		Gas:        21000,
+		BlobFeeCap: uint256.NewInt(params.GWei),
+		BlobHashes: blobHashes,
+		Value:      uint256.NewInt(100),
+		Sidecar:    types.NewBlobTxSidecar(version, blobs, commitments, proofs),
+	}
+	return types.MustSignNewTx(key, types.LatestSigner(chainConfig), blobtx)
+}
+
+// newBenchmarkBlobEnv creates an environment for blob benchmarks.
+// It creates multiple keys and fills the pool with blob transactions totaling the specified blob count.
+// version: 0 = BlobSidecarVersion0 (pre-Osaka), 1 = BlobSidecarVersion1 (Osaka+)
+// fork: which fork to enable
+func newBenchmarkBlobEnv(b *testing.B, blobCount int, version byte, fork benchFork) *benchmarkBlobEnv {
+	// Create a configuration that allows enough blobs
+	config := *params.MergedTestChainConfig
+	// Set blob schedule to allow for large blob counts (up to 128 blobs per block)
+	config.BlobScheduleConfig = &params.BlobScheduleConfig{
+		Cancun: &params.BlobConfig{Target: 6, Max: 128, UpdateFraction: 3338477},
+		Prague: &params.BlobConfig{Target: 6, Max: 128, UpdateFraction: 5007716},
+		Osaka:  &params.BlobConfig{Target: 6, Max: 128, UpdateFraction: 5007716},
+	}
+	// Configure fork times based on requested fork
+	switch fork {
+	case forkCancun:
+		config.PragueTime = nil
+		config.OsakaTime = nil
+	case forkPrague:
+		config.OsakaTime = nil
+	case forkOsaka:
+		// All forks enabled (default)
+	}
+
+	// Generate enough keys for all the blob transactions
+	// Each tx can have up to 6 blobs, so we need ceil(blobCount/6) keys
+	numTxs := (blobCount + 5) / 6
+	keys := make([]*ecdsa.PrivateKey, numTxs)
+	addrs := make([]common.Address, numTxs)
+	alloc := make(types.GenesisAlloc)
+	alloc[testAddr] = types.Account{Balance: testBalance}
+
+	for i := 0; i < numTxs; i++ {
+		key, _ := crypto.GenerateKey()
+		keys[i] = key
+		addrs[i] = crypto.PubkeyToAddress(key.PublicKey)
+		// Give each account enough balance for many transactions
+		alloc[addrs[i]] = types.Account{Balance: new(big.Int).Mul(big.NewInt(1e18), big.NewInt(10000))}
+	}
+
+	gspec := &core.Genesis{
+		Config:     &config,
+		Alloc:      alloc,
+		Difficulty: common.Big0,
+	}
+	n, ethServ := startEthService(b, gspec, nil)
+
+	// Collect versioned hashes for the blobs we'll use
+	var vhashes []common.Hash
+	for i := 0; i < blobCount; i++ {
+		vhashes = append(vhashes, benchBlobVHashes[i])
+	}
+
+	// Fill initial blob txs into the pool
+	env := &benchmarkBlobEnv{
+		node:      n,
+		eth:       ethServ,
+		api:       newConsensusAPIWithoutHeartbeat(ethServ),
+		config:    &config,
+		keys:      keys,
+		vhashes:   vhashes,
+		version:   version,
+		blobCount: blobCount,
+		nonces:    make([]uint64, numTxs),
+	}
+	env.addBlobTxs(b)
+	return env
+}
+
+// addBlobTxs adds blob transactions to the pool using the stored blobCount and per-key nonces.
+// It increments each key's nonce after adding transactions.
+func (env *benchmarkBlobEnv) addBlobTxs(b *testing.B) {
+	numTxs := (env.blobCount + 5) / 6
+	var txs []*types.Transaction
+	blobsRemaining := env.blobCount
+	blobOffset := 0
+
+	for i := 0; i < numTxs && blobsRemaining > 0; i++ {
+		// Each tx gets up to 6 blobs
+		txBlobCount := 6
+		if blobsRemaining < 6 {
+			txBlobCount = blobsRemaining
+		}
+		tx := makeBenchBlobTx(env.config, env.nonces[i], txBlobCount, blobOffset, env.keys[i], env.version)
+		txs = append(txs, tx)
+		blobOffset += txBlobCount
+		blobsRemaining -= txBlobCount
+	}
+	errs := env.eth.TxPool().Add(txs, true)
+	for i, err := range errs {
+		if err != nil {
+			b.Fatalf("Failed to add blob tx %d to pool: %v", i, err)
+		}
+	}
+	// Increment nonce for each key used
+	for i := 0; i < numTxs; i++ {
+		env.nonces[i]++
+	}
+}
+
+// Close closes the environment
+func (env *benchmarkBlobEnv) Close() {
+	env.node.Close()
+}
+
+// BenchmarkGetBlobsV1 benchmarks the GetBlobsV1 method with various blob counts.
+// GetBlobsV1 is available at Cancun/Prague (pre-Osaka).
+func BenchmarkGetBlobsV1(b *testing.B) {
+	for _, blobCount := range benchmarkBlobCounts {
+		for _, enc := range encodingTypes {
+			b.Run(fmt.Sprintf("blobs=%d/enc=%s", blobCount, enc), func(b *testing.B) {
+				env := newBenchmarkBlobEnv(b, blobCount, 0, forkPrague)
+				defer env.Close()
+
+				b.ResetTimer()
+				for b.Loop() {
+					result, err := env.api.GetBlobsV1(env.vhashes)
+					if err != nil {
+						b.Fatalf("GetBlobsV1 failed: %v", err)
+					}
+					// Verify we got the expected number of blobs
+					if len(result) != blobCount {
+						b.Fatalf("expected %d blobs, got %d", blobCount, len(result))
+					}
+					benchEncode(b, enc, result)
+				}
+				b.ReportMetric(float64(b.Elapsed().Milliseconds())/float64(b.N), "ms/op")
+			})
+		}
+	}
+}
+
+// BenchmarkGetBlobsV2Extended benchmarks the GetBlobsV2 method with various blob counts.
+// GetBlobsV2 is available at Osaka+.
+func BenchmarkGetBlobsV2Extended(b *testing.B) {
+	for _, blobCount := range benchmarkBlobCounts {
+		for _, enc := range encodingTypes {
+			b.Run(fmt.Sprintf("blobs=%d/enc=%s", blobCount, enc), func(b *testing.B) {
+				env := newBenchmarkBlobEnv(b, blobCount, 1, forkOsaka)
+				defer env.Close()
+
+				b.ResetTimer()
+				for b.Loop() {
+					result, err := env.api.GetBlobsV2(env.vhashes)
+					if err != nil {
+						b.Fatalf("GetBlobsV2 failed: %v", err)
+					}
+					// Verify we got the expected number of blobs
+					if len(result) != blobCount {
+						b.Fatalf("expected %d blobs, got %d", blobCount, len(result))
+					}
+					benchEncode(b, enc, result)
+				}
+				b.ReportMetric(float64(b.Elapsed().Milliseconds())/float64(b.N), "ms/op")
+			})
+		}
+	}
+}
+
+// BenchmarkGetBlobsV3 benchmarks the GetBlobsV3 method with various blob counts.
+// GetBlobsV3 is available at Osaka+.
+func BenchmarkGetBlobsV3(b *testing.B) {
+	for _, blobCount := range benchmarkBlobCounts {
+		for _, enc := range encodingTypes {
+			b.Run(fmt.Sprintf("blobs=%d/enc=%s", blobCount, enc), func(b *testing.B) {
+				env := newBenchmarkBlobEnv(b, blobCount, 1, forkOsaka)
+				defer env.Close()
+
+				b.ResetTimer()
+				for b.Loop() {
+					result, err := env.api.GetBlobsV3(env.vhashes)
+					if err != nil {
+						b.Fatalf("GetBlobsV3 failed: %v", err)
+					}
+					// Verify we got the expected number of blobs
+					if len(result) != blobCount {
+						b.Fatalf("expected %d blobs, got %d", blobCount, len(result))
+					}
+					benchEncode(b, enc, result)
+				}
+				b.ReportMetric(float64(b.Elapsed().Milliseconds())/float64(b.N), "ms/op")
+			})
+		}
+	}
+}
+
+// BenchmarkGetPayloadV5WithBlobs benchmarks GetPayloadV5 (Osaka fork) with blobs.
+// Note: Measures single iteration performance due to NewPayload complexity at Osaka.
+func BenchmarkGetPayloadV5WithBlobs(b *testing.B) {
+	for _, blobCount := range benchmarkBlobCounts {
+		for _, enc := range encodingTypes {
+			b.Run(fmt.Sprintf("blobs=%d/enc=%s", blobCount, enc), func(b *testing.B) {
+				env := newBenchmarkBlobEnv(b, blobCount, 1, forkOsaka)
+				defer env.Close()
+
+				parent := env.api.eth.BlockChain().CurrentHeader()
+				beaconRoot := common.Hash{0x42}
+
+				b.ResetTimer()
+				for i := 0; i < b.N; i++ {
+					// Note: We don't call addBlobTxs here because we can't advance the chain
+					// (NewPayloadV5 requires execution requests). The same transactions are
+					// reused for each iteration, which still benchmarks the GetPayload performance.
+					timestamp := parent.Time + 12
+					fcState := engine.ForkchoiceStateV1{
+						HeadBlockHash:      parent.Hash(),
+						SafeBlockHash:      parent.Hash(),
+						FinalizedBlockHash: parent.Hash(),
+					}
+					payloadAttr := &engine.PayloadAttributes{
+						Timestamp:             timestamp,
+						Random:                common.Hash{byte(i)},
+						SuggestedFeeRecipient: testAddr,
+						Withdrawals:           []*types.Withdrawal{},
+						BeaconRoot:            &beaconRoot,
+					}
+					resp, err := env.api.ForkchoiceUpdatedV3(context.Background(), fcState, payloadAttr)
+					if err != nil {
+						b.Fatalf("ForkchoiceUpdatedV3 failed: %v", err)
+					}
+					if resp.PayloadID == nil {
+						b.Fatalf("ForkchoiceUpdatedV3 returned nil PayloadID")
+					}
+					// Wait for the payload to be built with transactions
+					time.Sleep(100 * time.Millisecond)
+
+					envelope, err := env.api.GetPayloadV5(*resp.PayloadID)
+					if err != nil {
+						b.Fatalf("GetPayloadV5 failed: %v", err)
+					}
+					if envelope.BlobsBundle == nil {
+						b.Fatalf("BlobsBundle is nil")
+					}
+					// Verify we got the expected number of blobs
+					if len(envelope.BlobsBundle.Blobs) != blobCount {
+						b.Fatalf("expected %d blobs, got %d", blobCount, len(envelope.BlobsBundle.Blobs))
+					}
+					benchEncode(b, enc, envelope)
+				}
+				b.ReportMetric(float64(b.Elapsed().Milliseconds())/float64(b.N), "ms/op")
+			})
+		}
+	}
+}
+
+// BenchmarkNewPayloadV3WithBlobs benchmarks the NewPayloadV3 method with various blob counts.
+// Each iteration processes a payload with the full blob count.
+func BenchmarkNewPayloadV3WithBlobs(b *testing.B) {
+	for _, blobCount := range benchmarkBlobCounts {
+		for _, enc := range encodingTypes {
+			b.Run(fmt.Sprintf("blobs=%d/enc=%s", blobCount, enc), func(b *testing.B) {
+				env := newBenchmarkBlobEnv(b, blobCount, 0, forkCancun)
+				defer env.Close()
+
+				parent := env.api.eth.BlockChain().CurrentHeader()
+				beaconRoot := common.Hash{0x42}
+
+				// Build a payload first to get valid executable data
+				timestamp := parent.Time + 12
+				fcState := engine.ForkchoiceStateV1{
+					HeadBlockHash:      parent.Hash(),
+					SafeBlockHash:      parent.Hash(),
+					FinalizedBlockHash: parent.Hash(),
+				}
+				payloadAttr := &engine.PayloadAttributes{
+					Timestamp:             timestamp,
+					Random:                common.Hash{0x01},
+					SuggestedFeeRecipient: testAddr,
+					Withdrawals:           []*types.Withdrawal{},
+					BeaconRoot:            &beaconRoot,
+				}
+				resp, err := env.api.ForkchoiceUpdatedV3(context.Background(), fcState, payloadAttr)
+				if err != nil {
+					b.Fatalf("ForkchoiceUpdatedV3 failed: %v", err)
+				}
+				if resp.PayloadID == nil {
+					b.Fatalf("ForkchoiceUpdatedV3 returned nil PayloadID")
+				}
+				// Wait for the payload to be built with transactions
+				time.Sleep(100 * time.Millisecond)
+
+				// Get the payload
+				envelope, err := env.api.GetPayloadV3(*resp.PayloadID)
+				if err != nil {
+					b.Fatalf("GetPayloadV3 failed: %v", err)
+				}
+				// Verify we got the expected number of blobs
+				if len(envelope.BlobsBundle.Blobs) != blobCount {
+					b.Fatalf("expected %d blobs in setup, got %d", blobCount, len(envelope.BlobsBundle.Blobs))
+				}
+
+				execData := envelope.ExecutionPayload
+				// Collect versioned hashes from blobs bundle
+				vhashes := make([]common.Hash, len(envelope.BlobsBundle.Commitments))
+				for j, commitment := range envelope.BlobsBundle.Commitments {
+					var commit kzg4844.Commitment
+					copy(commit[:], commitment)
+					vhashes[j] = kzg4844.CalcBlobHashV1(sha256.New(), &commit)
+				}
+
+				b.ResetTimer()
+				for i := 0; i < b.N; i++ {
+					// NewPayload is idempotent, calling it multiple times with the same data
+					// should return the same result. The payload contains blobCount blobs.
+					result, err := env.api.NewPayloadV3(context.Background(), *execData, vhashes, &beaconRoot)
+					if err != nil {
+						b.Fatalf("NewPayloadV3 failed: %v", err)
+					}
+					benchEncode(b, enc, result)
+				}
+				b.ReportMetric(float64(b.Elapsed().Milliseconds())/float64(b.N), "ms/op")
+			})
+		}
+	}
+}
+
+// BenchmarkForkchoiceUpdatedWithBlobPayload benchmarks forkchoice updates that trigger
+// payload building with blob transactions.
+// Note: Measures ForkchoiceUpdated performance with blob transactions in the pool.
+func BenchmarkForkchoiceUpdatedWithBlobPayload(b *testing.B) {
+	for _, blobCount := range benchmarkBlobCounts {
+		for _, enc := range encodingTypes {
+			b.Run(fmt.Sprintf("blobs=%d/enc=%s", blobCount, enc), func(b *testing.B) {
+				env := newBenchmarkBlobEnv(b, blobCount, 0, forkCancun)
+				defer env.Close()
+
+				parent := env.api.eth.BlockChain().CurrentHeader()
+				beaconRoot := common.Hash{0x42}
+
+				b.ResetTimer()
+				for i := 0; i < b.N; i++ {
+					// Note: We don't call addBlobTxs here because the blob pool has
+					// a per-account limit of 16 transactions. The same transactions are
+					// reused for each iteration, which still benchmarks the ForkchoiceUpdated
+					// performance with blob transactions in the pool.
+					timestamp := parent.Time + 12
+					fcState := engine.ForkchoiceStateV1{
+						HeadBlockHash:      parent.Hash(),
+						SafeBlockHash:      parent.Hash(),
+						FinalizedBlockHash: parent.Hash(),
+					}
+					payloadAttr := &engine.PayloadAttributes{
+						Timestamp:             timestamp,
+						Random:                common.Hash{byte(i)},
+						SuggestedFeeRecipient: testAddr,
+						Withdrawals:           []*types.Withdrawal{},
+						BeaconRoot:            &beaconRoot,
+					}
+					resp, err := env.api.ForkchoiceUpdatedV3(context.Background(), fcState, payloadAttr)
+					if err != nil {
+						b.Fatalf("ForkchoiceUpdatedV3 failed: %v", err)
+					}
+					if resp.PayloadID == nil {
+						b.Fatalf("ForkchoiceUpdatedV3 returned nil PayloadID")
+					}
+					benchEncode(b, enc, resp)
+				}
+				b.ReportMetric(float64(b.Elapsed().Milliseconds())/float64(b.N), "ms/op")
+			})
+		}
+	}
+}
+
+// BenchmarkFullBlobWorkflowOsaka benchmarks the complete blob workflow at Osaka:
+// ForkchoiceUpdated -> GetPayload
+// Note: Measures single iteration performance due to NewPayload complexity at Osaka.
+func BenchmarkFullBlobWorkflowOsaka(b *testing.B) {
+	for _, blobCount := range benchmarkBlobCounts {
+		for _, enc := range encodingTypes {
+			b.Run(fmt.Sprintf("blobs=%d/enc=%s", blobCount, enc), func(b *testing.B) {
+				env := newBenchmarkBlobEnv(b, blobCount, 1, forkOsaka)
+				defer env.Close()
+
+				parent := env.api.eth.BlockChain().CurrentHeader()
+				beaconRoot := common.Hash{0x42}
+
+				b.ResetTimer()
+				for i := 0; i < b.N; i++ {
+					// Note: We don't call addBlobTxs here because we can't advance the chain
+					// (NewPayloadV5 requires execution requests). The same transactions are
+					// reused for each iteration, which still benchmarks the workflow performance.
+
+					// 1. ForkchoiceUpdated to build payload
+					timestamp := parent.Time + 12
+					fcState := engine.ForkchoiceStateV1{
+						HeadBlockHash:      parent.Hash(),
+						SafeBlockHash:      parent.Hash(),
+						FinalizedBlockHash: parent.Hash(),
+					}
+					payloadAttr := &engine.PayloadAttributes{
+						Timestamp:             timestamp,
+						Random:                common.Hash{byte(i)},
+						SuggestedFeeRecipient: testAddr,
+						Withdrawals:           []*types.Withdrawal{},
+						BeaconRoot:            &beaconRoot,
+					}
+					resp, err := env.api.ForkchoiceUpdatedV3(context.Background(), fcState, payloadAttr)
+					if err != nil {
+						b.Fatalf("ForkchoiceUpdatedV3 failed: %v", err)
+					}
+					if resp.PayloadID == nil {
+						b.Fatalf("ForkchoiceUpdatedV3 returned nil PayloadID")
+					}
+					// Encode ForkchoiceUpdated response
+					benchEncode(b, enc, resp)
+
+					// Wait for the payload to be built with transactions
+					time.Sleep(100 * time.Millisecond)
+
+					// 2. GetPayload
+					envelope, err := env.api.GetPayloadV5(*resp.PayloadID)
+					if err != nil {
+						b.Fatalf("GetPayloadV5 failed: %v", err)
+					}
+					if envelope.BlobsBundle == nil {
+						b.Fatalf("BlobsBundle is nil")
+					}
+					// Verify we got the expected number of blobs
+					if len(envelope.BlobsBundle.Blobs) != blobCount {
+						b.Fatalf("expected %d blobs, got %d", blobCount, len(envelope.BlobsBundle.Blobs))
+					}
+					// Encode GetPayload response
+					benchEncode(b, enc, envelope)
+				}
+				b.ReportMetric(float64(b.Elapsed().Milliseconds())/float64(b.N), "ms/op")
+			})
+		}
+	}
+}
+
+// discardConn is a net.Conn-like writer that discards all output.
+// Used to measure server-side RPC cost without client-side decoding.
+type discardConn struct {
+	io.Reader
+	io.Writer
+}
+
+func (discardConn) Close() error                     { return nil }
+func (discardConn) SetWriteDeadline(time.Time) error { return nil }
+
+// BenchmarkGetPayloadV5RPCServerOnly benchmarks only the EL server-side cost of
+// engine_getPayloadV5: method dispatch, JSON serialization, and wire encoding.
+// Client-side decoding is excluded by writing to io.Discard.
+func BenchmarkGetPayloadV5RPCServerOnly(b *testing.B) {
+	blobCount := 72
+	env := newBenchmarkBlobEnv(b, blobCount, 1, forkOsaka)
+	defer env.Close()
+
+	// Register the engine API on the running node's in-process RPC server.
+	rpcServer, err := env.node.RPCHandler()
+	if err != nil {
+		b.Fatalf("RPCHandler failed: %v", err)
+	}
+	rpcServer.RegisterName("engine", env.api)
+
+	parent := env.api.eth.BlockChain().CurrentHeader()
+	beaconRoot := common.Hash{0x42}
+
+	// Build one payload to get a valid payloadID.
+	fcState := engine.ForkchoiceStateV1{
+		HeadBlockHash:      parent.Hash(),
+		SafeBlockHash:      parent.Hash(),
+		FinalizedBlockHash: parent.Hash(),
+	}
+	payloadAttr := &engine.PayloadAttributes{
+		Timestamp:             parent.Time + 12,
+		Random:                common.Hash{0x01},
+		SuggestedFeeRecipient: testAddr,
+		Withdrawals:           []*types.Withdrawal{},
+		BeaconRoot:            &beaconRoot,
+	}
+	resp, err := env.api.ForkchoiceUpdatedV3(context.Background(), fcState, payloadAttr)
+	if err != nil {
+		b.Fatalf("ForkchoiceUpdatedV3 failed: %v", err)
+	}
+	if resp.PayloadID == nil {
+		b.Fatalf("ForkchoiceUpdatedV3 returned nil PayloadID")
+	}
+	time.Sleep(100 * time.Millisecond)
+
+	// Verify the payload has the expected blobs via the direct API first.
+	envelope, err := env.api.GetPayloadV5(*resp.PayloadID)
+	if err != nil {
+		b.Fatalf("GetPayloadV5 failed: %v", err)
+	}
+	if len(envelope.BlobsBundle.Blobs) != blobCount {
+		b.Fatalf("expected %d blobs, got %d", blobCount, len(envelope.BlobsBundle.Blobs))
+	}
+	b.Logf("payload size: %d blobs, %d txs", len(envelope.BlobsBundle.Blobs), len(envelope.ExecutionPayload.Transactions))
+
+	// Build the JSON-RPC request bytes once.
+	reqJSON := fmt.Sprintf(`{"jsonrpc":"2.0","id":1,"method":"engine_getPayloadV5","params":["%s"]}`, resp.PayloadID.String())
+
+	b.ResetTimer()
+	for i := 0; i < b.N; i++ {
+		conn := discardConn{
+			Reader: strings.NewReader(reqJSON),
+			Writer: io.Discard,
+		}
+		codec := rpc.NewCodec(conn)
+		rpcServer.ServeCodec(codec, 0)
+	}
+	b.StopTimer()
+	b.ReportMetric(float64(b.Elapsed().Milliseconds())/float64(b.N), "ms/op")
+}
+
+// BenchmarkGetBlobsV3RPCServerOnly benchmarks only the EL server-side cost of
+// engine_getBlobsV3: method dispatch, JSON serialization, and wire encoding.
+// Client-side decoding is excluded by writing to io.Discard.
+func BenchmarkGetBlobsV3RPCServerOnly(b *testing.B) {
+	blobCount := 72
+	env := newBenchmarkBlobEnv(b, blobCount, 1, forkOsaka)
+	defer env.Close()
+
+	// Register the engine API on the running node's in-process RPC server.
+	rpcServer, err := env.node.RPCHandler()
+	if err != nil {
+		b.Fatalf("RPCHandler failed: %v", err)
+	}
+	rpcServer.RegisterName("engine", env.api)
+
+	// Verify the blobs are available via the direct API first.
+	result, err := env.api.GetBlobsV3(env.vhashes)
+	if err != nil {
+		b.Fatalf("GetBlobsV3 failed: %v", err)
+	}
+	if len(result) != blobCount {
+		b.Fatalf("expected %d blobs, got %d", blobCount, len(result))
+	}
+	b.Logf("blob count: %d", blobCount)
+
+	// Build the JSON-RPC request bytes once.
+	// Format the versioned hashes as a JSON array of hex strings.
+	var hashStrs []string
+	for _, h := range env.vhashes {
+		hashStrs = append(hashStrs, fmt.Sprintf(`"%s"`, h.Hex()))
+	}
+	reqJSON := fmt.Sprintf(`{"jsonrpc":"2.0","id":1,"method":"engine_getBlobsV3","params":[[%s]]}`, strings.Join(hashStrs, ","))
+
+	b.ResetTimer()
+	for i := 0; i < b.N; i++ {
+		conn := discardConn{
+			Reader: strings.NewReader(reqJSON),
+			Writer: io.Discard,
+		}
+		codec := rpc.NewCodec(conn)
+		rpcServer.ServeCodec(codec, 0)
+	}
+	b.StopTimer()
+	b.ReportMetric(float64(b.Elapsed().Milliseconds())/float64(b.N), "ms/op")
+}

eth/catalyst/api_test.go

@@ -1961,7 +1961,7 @@ func TestGetBlobsV1(t *testing.T) {
 		// Fill the request for retrieving blobs
 		var (
 			vhashes []common.Hash
-			expect  []*engine.BlobAndProofV1
+			expect  engine.BlobAndProofListV1
 		)
 		// fill missing blob at the beginning
 		if suite.fillRandom {
@@ -2072,13 +2072,13 @@ func BenchmarkGetBlobsV2(b *testing.B) {
 	}
 }
 
-type getBlobsFn func(hashes []common.Hash) ([]*engine.BlobAndProofV2, error)
+type getBlobsFn func(hashes []common.Hash) (engine.BlobAndProofListV2, error)
 
 func runGetBlobs(t testing.TB, getBlobs getBlobsFn, start, limit int, fillRandom bool, expectPartialResponse bool, name string) {
 	// Fill the request for retrieving blobs
 	var (
 		vhashes []common.Hash
-		expect  []*engine.BlobAndProofV2
+		expect  engine.BlobAndProofListV2
 	)
 	for j := start; j < limit; j++ {
 		vhashes = append(vhashes, testBlobVHashes[j])

go.mod

@@ -26,6 +26,7 @@ require (
 	github.com/ethereum/hid v1.0.1-0.20260421154323-c2ab8d9bf68a
 	github.com/fatih/color v1.16.0
 	github.com/ferranbt/fastssz v0.1.4
+	github.com/fjl/jsonw v0.1.0
 	github.com/fsnotify/fsnotify v1.6.0
 	github.com/gballet/go-libpcsclite v0.0.0-20190607065134-2772fd86a8ff
 	github.com/gofrs/flock v0.12.1
@@ -121,7 +122,7 @@ require (
 	github.com/deepmap/oapi-codegen v1.6.0 // indirect
 	github.com/dlclark/regexp2 v1.7.0 // indirect
 	github.com/emicklei/dot v1.6.2 // indirect
-	github.com/fjl/gencodec v0.1.0 // indirect
+	github.com/fjl/gencodec v0.1.2 // indirect
 	github.com/garslo/gogen v0.0.0-20170306192744-1d203ffc1f61 // indirect
 	github.com/getsentry/sentry-go v0.27.0 // indirect
 	github.com/go-ole/go-ole v1.3.0 // indirect

go.sum

@@ -123,8 +123,10 @@ github.com/fatih/color v1.16.0 h1:zmkK9Ngbjj+K0yRhTVONQh1p/HknKYSlNT+vZCzyokM=
 github.com/fatih/color v1.16.0/go.mod h1:fL2Sau1YI5c0pdGEVCbKQbLXB6edEj1ZgiY4NijnWvE=
 github.com/ferranbt/fastssz v0.1.4 h1:OCDB+dYDEQDvAgtAGnTSidK1Pe2tW3nFV40XyMkTeDY=
 github.com/ferranbt/fastssz v0.1.4/go.mod h1:Ea3+oeoRGGLGm5shYAeDgu6PGUlcvQhE2fILyD9+tGg=
-github.com/fjl/gencodec v0.1.0 h1:B3K0xPfc52cw52BBgUbSPxYo+HlLfAgWMVKRWXUXBcs=
-github.com/fjl/gencodec v0.1.0/go.mod h1:Um1dFHPONZGTHog1qD1NaWjXJW/SPB38wPv0O8uZ2fI=
+github.com/fjl/gencodec v0.1.2 h1:nf+MMsmuii5ZQMbS6/xjZoe5LRkN0415FOJOSwmnuW8=
+github.com/fjl/gencodec v0.1.2/go.mod h1:chDHL3wKXuBgauP8x3XNZkl5EIAR5SoCTmmmDTZRzmw=
+github.com/fjl/jsonw v0.1.0 h1:V3MyR79fjLpn/+bMgvegdGUIhoJOzjmqWcKDgcOmY1I=
+github.com/fjl/jsonw v0.1.0/go.mod h1:2KMLevM6FXEJnfhtk7naXu9vZdVfOma1GlnGdPRlumU=
 github.com/fsnotify/fsnotify v1.4.7/go.mod h1:jwhsz4b93w/PPRr/qN1Yymfu8t87LnFCMoQvtojpjFo=
 github.com/fsnotify/fsnotify v1.4.9/go.mod h1:znqG4EE+3YCdAaPaxE2ZRY/06pZUdp0tY4IgpuI1SZQ=
 github.com/fsnotify/fsnotify v1.6.0 h1:n+5WquG0fcWoWp6xPWfHdbskMCQaFnG6PfBrh1Ky4HY=

rpc/client.go

@@ -364,7 +364,7 @@ func (c *Client) CallContext(ctx context.Context, result interface{}, method str
 	resp := batchresp[0]
 	switch {
 	case resp.Error != nil:
-		return resp.Error
+		return resp.decodeError()
 	case len(resp.Result) == 0:
 		return ErrNoResult
 	default:
@@ -419,7 +419,7 @@ func (c *Client) BatchCallContext(ctx context.Context, b []BatchElem) error {
 	if c.isHTTP {
 		err = c.sendBatchHTTP(ctx, op, msgs)
 	} else {
-		err = c.send(ctx, op, msgs)
+		err = c.sendBatch(ctx, op, msgs)
 	}
 	if err != nil {
 		return err
@@ -449,7 +449,7 @@ func (c *Client) BatchCallContext(ctx context.Context, b []BatchElem) error {
 		elem := &b[index]
 		switch {
 		case resp.Error != nil:
-			elem.Error = resp.Error
+			elem.Error = resp.decodeError()
 		case resp.Result == nil:
 			elem.Error = ErrNoResult
 		default:
@@ -552,7 +552,7 @@ func (c *Client) newMessage(method string, paramsIn ...interface{}) (*jsonrpcMes
 
 // send registers op with the dispatch loop, then sends msg on the connection.
 // if sending fails, op is deregistered.
-func (c *Client) send(ctx context.Context, op *requestOp, msg interface{}) error {
+func (c *Client) send(ctx context.Context, op *requestOp, msg *jsonrpcMessage) error {
 	select {
 	case c.reqInit <- op:
 		err := c.write(ctx, msg, false)
@@ -567,7 +567,22 @@ func (c *Client) send(ctx context.Context, op *requestOp, msg interface{}) error
 	}
 }
 
-func (c *Client) write(ctx context.Context, msg interface{}, retry bool) error {
+// sendBatch registers op with the dispatch loop, then sends a batch of messages
+// on the connection. If sending fails, op is deregistered.
+func (c *Client) sendBatch(ctx context.Context, op *requestOp, msgs []*jsonrpcMessage) error {
+	select {
+	case c.reqInit <- op:
+		err := c.writeBatch(ctx, msgs, false)
+		c.reqSent <- err
+		return err
+	case <-ctx.Done():
+		return ctx.Err()
+	case <-c.closing:
+		return ErrClientQuit
+	}
+}
+
+func (c *Client) write(ctx context.Context, msg *jsonrpcMessage, retry bool) error {
 	if c.writeConn == nil {
 		// The previous write failed. Try to establish a new connection.
 		if err := c.reconnect(ctx); err != nil {
@@ -584,6 +599,22 @@ func (c *Client) write(ctx context.Context, msg interface{}, retry bool) error {
 	return err
 }
 
+func (c *Client) writeBatch(ctx context.Context, msgs []*jsonrpcMessage, retry bool) error {
+	if c.writeConn == nil {
+		if err := c.reconnect(ctx); err != nil {
+			return err
+		}
+	}
+	err := c.writeConn.writeJSONBatch(ctx, msgs, false)
+	if err != nil {
+		c.writeConn = nil
+		if !retry {
+			return c.writeBatch(ctx, msgs, true)
+		}
+	}
+	return err
+}
+
 func (c *Client) reconnect(ctx context.Context) error {
 	if c.reconnectFunc == nil {
 		return errDead

rpc/handler.go

@@ -169,7 +169,7 @@ func (b *batchCallBuffer) doWrite(ctx context.Context, conn jsonWriter, isErrorR
 	}
 	b.wrote = true // can only write once
 	if len(b.resp) > 0 {
-		conn.writeJSON(ctx, b.resp, isErrorResponse)
+		conn.writeJSONBatch(ctx, b.resp, isErrorResponse)
 	}
 }
 
@@ -237,7 +237,7 @@ func (h *handler) handleBatch(msgs []*jsonrpcMessage) {
 			resp := h.handleCallMsg(cp, msg)
 			callBuffer.pushResponse(resp)
 			if resp != nil && h.batchResponseMaxSize != 0 {
-				responseBytes += len(resp.Result)
+				responseBytes += len(resp.Result) + len(resp.Error)
 				if responseBytes > h.batchResponseMaxSize {
 					err := &internalServerError{errcodeResponseTooLarge, errMsgResponseTooLarge}
 					callBuffer.respondWithError(cp.ctx, h.conn, err)
@@ -268,7 +268,7 @@ func (h *handler) respondWithBatchTooLarge(cp *callProc, batch []*jsonrpcMessage
 			break
 		}
 	}
-	h.conn.writeJSON(cp.ctx, []*jsonrpcMessage{resp}, true)
+	h.conn.writeJSONBatch(cp.ctx, []*jsonrpcMessage{resp}, true)
 }
 
 // handleMsg handles a single non-batch message.
@@ -415,7 +415,7 @@ func (h *handler) handleResponses(batch []*jsonrpcMessage, handleCall func(*json
 		// the op.resp channel.
 		if op.sub != nil {
 			if msg.Error != nil {
-				op.err = msg.Error
+				op.err = msg.decodeError()
 			} else {
 				op.err = json.Unmarshal(msg.Result, &op.sub.subid)
 				if op.err == nil {
@@ -481,9 +481,10 @@ func (h *handler) handleCallMsg(ctx *callProc, msg *jsonrpcMessage) *jsonrpcMess
 		var logctx []any
 		logctx = append(logctx, "reqid", idForLog{msg.ID}, "duration", time.Since(start))
 		if resp.Error != nil {
-			logctx = append(logctx, "err", resp.Error.Message)
-			if resp.Error.Data != nil {
-				logctx = append(logctx, "errdata", formatErrorData(resp.Error.Data))
+			je := resp.decodeError()
+			logctx = append(logctx, "err", je.Message)
+			if je.Data != nil {
+				logctx = append(logctx, "errdata", formatErrorData(je.Data))
 			}
 			h.log.Warn("Served "+msg.Method, logctx...)
 		} else {
@@ -550,7 +551,7 @@ func (h *handler) handleCall(cp *callProc, msg *jsonrpcMessage) *jsonrpcMessage
 	answer := h.runMethod(rctx, msg, callb, args)
 	var rErr error
 	if answer.Error != nil {
-		rErr = errors.New(answer.Error.Message)
+		rErr = errors.New(answer.decodeError().Message)
 	}
 	rSpanEnd(&rErr)
 
@@ -623,7 +624,7 @@ func (h *handler) runMethod(ctx context.Context, msg *jsonrpcMessage, callb *cal
 	_, _, spanEnd := telemetry.StartSpanWithTracer(ctx, h.tracer(), "rpc.encodeJSONResponse", attributes...)
 	response := msg.response(result)
 	if response.Error != nil {
-		err = errors.New(response.Error.Message)
+		err = errors.New(response.decodeError().Message)
 	}
 	spanEnd(&err)
 	return response

rpc/http.go

@@ -57,10 +57,14 @@ type httpConn struct {
 // and some methods don't work. The panic() stubs here exist to ensure
 // this special treatment is correct.
 
-func (hc *httpConn) writeJSON(context.Context, interface{}, bool) error {
+func (hc *httpConn) writeJSON(context.Context, *jsonrpcMessage, bool) error {
 	panic("writeJSON called on httpConn")
 }
 
+func (hc *httpConn) writeJSONBatch(context.Context, []*jsonrpcMessage, bool) error {
+	panic("writeJSONBatch called on httpConn")
+}
+
 func (hc *httpConn) peerInfo() PeerInfo {
 	panic("peerInfo called on httpConn")
 }
@@ -179,9 +183,9 @@ func cleanlyCloseBody(body io.ReadCloser) error {
 	return body.Close()
 }
 
-func (c *Client) sendHTTP(ctx context.Context, op *requestOp, msg interface{}) error {
+func (c *Client) sendHTTP(ctx context.Context, op *requestOp, msg *jsonrpcMessage) error {
 	hc := c.writeConn.(*httpConn)
-	respBody, err := hc.doRequest(ctx, msg)
+	respBody, err := hc.doRequest(ctx, appendMessage(nil, msg))
 	if err != nil {
 		return err
 	}
@@ -198,7 +202,7 @@ func (c *Client) sendHTTP(ctx context.Context, op *requestOp, msg interface{}) e
 
 func (c *Client) sendBatchHTTP(ctx context.Context, op *requestOp, msgs []*jsonrpcMessage) error {
 	hc := c.writeConn.(*httpConn)
-	respBody, err := hc.doRequest(ctx, msgs)
+	respBody, err := hc.doRequest(ctx, appendBatch(nil, msgs))
 	if err != nil {
 		return err
 	}
@@ -212,11 +216,7 @@ func (c *Client) sendBatchHTTP(ctx context.Context, op *requestOp, msgs []*jsonr
 	return nil
 }
 
-func (hc *httpConn) doRequest(ctx context.Context, msg interface{}) (io.ReadCloser, error) {
-	body, err := json.Marshal(msg)
-	if err != nil {
-		return nil, err
-	}
+func (hc *httpConn) doRequest(ctx context.Context, body []byte) (io.ReadCloser, error) {
 	req, err := http.NewRequestWithContext(ctx, http.MethodPost, hc.url, io.NopCloser(bytes.NewReader(body)))
 	if err != nil {
 		return nil, err
@@ -268,41 +268,51 @@ func (s *Server) newHTTPServerConn(r *http.Request, w http.ResponseWriter) Serve
 	body := io.LimitReader(r.Body, int64(s.httpBodyLimit))
 	conn := &httpServerConn{Reader: body, Writer: w, r: r}
 
-	encoder := func(v any, isErrorResponse bool) error {
-		if !isErrorResponse {
-			return json.NewEncoder(conn).Encode(v)
-		}
-
-		// It's an error response and requires special treatment.
-		//
-		// In case of a timeout error, the response must be written before the HTTP
-		// server's write timeout occurs. So we need to flush the response. The
-		// Content-Length header also needs to be set to ensure the client knows
-		// when it has the full response.
-		encdata, err := json.Marshal(v)
-		if err != nil {
-			return err
-		}
-		w.Header().Set("content-length", strconv.Itoa(len(encdata)))
-
-		// If this request is wrapped in a handler that might remove Content-Length (such
-		// as the automatic gzip we do in package node), we need to ensure the HTTP server
-		// doesn't perform chunked encoding. In case WriteTimeout is reached, the chunked
-		// encoding might not be finished correctly, and some clients do not like it when
-		// the final chunk is missing.
-		w.Header().Set("transfer-encoding", "identity")
-
-		_, err = w.Write(encdata)
-		if f, ok := w.(http.Flusher); ok {
-			f.Flush()
-		}
-		return err
+	var buf []byte
+	encodeMsg := func(msg *jsonrpcMessage, isError bool) error {
+		buf = appendMessage(buf[:0], msg)
+		return httpWriteResult(w, buf, isError)
+	}
+	encodeBatch := func(msgs []*jsonrpcMessage, isError bool) error {
+		buf = appendBatch(buf[:0], msgs)
+		return httpWriteResult(w, buf, isError)
 	}
 
 	dec := json.NewDecoder(conn)
 	dec.UseNumber()
 
-	return NewFuncCodec(conn, encoder, dec.Decode)
+	return NewFuncCodec(conn, encodeMsg, encodeBatch, dec.Decode)
+}
+
+// httpWriteResult writes pre-encoded response data over HTTP.
+// For error responses, it sets Content-Length and flushes to ensure the response
+// is fully written before any HTTP server write timeout occurs.
+func httpWriteResult(w http.ResponseWriter, data []byte, isError bool) error {
+	if !isError {
+		_, err := w.Write(data)
+		return err
+	}
+
+	// It's an error response and requires special treatment.
+	//
+	// In case of a timeout error, the response must be written before the HTTP
+	// server's write timeout occurs. So we need to flush the response. The
+	// Content-Length header also needs to be set to ensure the client knows
+	// when it has the full response.
+	w.Header().Set("content-length", strconv.Itoa(len(data)))
+
+	// If this request is wrapped in a handler that might remove Content-Length (such
+	// as the automatic gzip we do in package node), we need to ensure the HTTP server
+	// doesn't perform chunked encoding. In case WriteTimeout is reached, the chunked
+	// encoding might not be finished correctly, and some clients do not like it when
+	// the final chunk is missing.
+	w.Header().Set("transfer-encoding", "identity")
+
+	_, err := w.Write(data)
+	if f, ok := w.(http.Flusher); ok {
+		f.Flush()
+	}
+	return err
 }
 
 // Close does nothing and always returns nil.

rpc/json.go

@@ -27,6 +27,8 @@ import (
 	"strings"
 	"sync"
 	"time"
+
+	"github.com/fjl/jsonw"
 )
 
 const (
@@ -52,12 +54,6 @@ type subscriptionResultEnc struct {
 	Result any    `json:"result"`
 }
 
-type jsonrpcSubscriptionNotification struct {
-	Version string                `json:"jsonrpc"`
-	Method  string                `json:"method"`
-	Params  subscriptionResultEnc `json:"params"`
-}
-
 // A value of this type can a JSON-RPC request, notification, successful response or
 // error response. Which one it is depends on the fields.
 type jsonrpcMessage struct {
@@ -65,7 +61,7 @@ type jsonrpcMessage struct {
 	ID      json.RawMessage `json:"id,omitempty"`
 	Method  string          `json:"method,omitempty"`
 	Params  json.RawMessage `json:"params,omitempty"`
-	Error   *jsonError      `json:"error,omitempty"`
+	Error   json.RawMessage `json:"error,omitempty"`
 	Result  json.RawMessage `json:"result,omitempty"`
 }
 
@@ -113,8 +109,29 @@ func (msg *jsonrpcMessage) errorResponse(err error) *jsonrpcMessage {
 	return resp
 }
 
+// decodeError decodes the Error field into a jsonError struct.
+func (msg *jsonrpcMessage) decodeError() *jsonError {
+	if msg.Error == nil {
+		return nil
+	}
+	je := new(jsonError)
+	json.Unmarshal(msg.Error, je)
+	return je
+}
+
 func (msg *jsonrpcMessage) response(result interface{}) *jsonrpcMessage {
-	enc, err := json.Marshal(result)
+	var (
+		enc []byte
+		err error
+	)
+	// Call MarshalJSON directly for types that implement it. This avoids the
+	// expensive validation/compaction pass that json.Marshal performs on
+	// encoder output.
+	if m, ok := result.(json.Marshaler); ok {
+		enc, err = m.MarshalJSON()
+	} else {
+		enc, err = json.Marshal(result)
+	}
 	if err != nil {
 		return msg.errorResponse(&internalServerError{errcodeMarshalError, err.Error()})
 	}
@@ -122,19 +139,18 @@ func (msg *jsonrpcMessage) response(result interface{}) *jsonrpcMessage {
 }
 
 func errorMessage(err error) *jsonrpcMessage {
-	msg := &jsonrpcMessage{Version: vsn, ID: null, Error: &jsonError{
+	je := &jsonError{
 		Code:    errcodeDefault,
 		Message: err.Error(),
-	}}
-	ec, ok := err.(Error)
-	if ok {
-		msg.Error.Code = ec.ErrorCode()
 	}
-	de, ok := err.(DataError)
-	if ok {
-		msg.Error.Data = de.ErrorData()
+	if ec, ok := err.(Error); ok {
+		je.Code = ec.ErrorCode()
 	}
-	return msg
+	if de, ok := err.(DataError); ok {
+		je.Data = de.ErrorData()
+	}
+	enc, _ := json.Marshal(je)
+	return &jsonrpcMessage{Version: vsn, ID: null, Error: enc}
 }
 
 type jsonError struct {
@@ -179,28 +195,32 @@ type ConnRemoteAddr interface {
 // jsonCodec reads and writes JSON-RPC messages to the underlying connection. It also has
 // support for parsing arguments and serializing (result) objects.
 type jsonCodec struct {
-	remote  string
-	closer  sync.Once        // close closed channel once
-	closeCh chan interface{} // closed on Close
-	decode  decodeFunc       // decoder to allow multiple transports
-	encMu   sync.Mutex       // guards the encoder
-	encode  encodeFunc       // encoder to allow multiple transports
-	conn    deadlineCloser
+	remote      string
+	closer      sync.Once        // close closed channel once
+	closeCh     chan interface{} // closed on Close
+	decode      decodeFunc       // decoder to allow multiple transports
+	encMu       sync.Mutex       // guards the encoder
+	encodeMsg   encodeMsgFunc    // single-message encoder
+	encodeBatch encodeBatchFunc  // batch encoder
+	conn        deadlineCloser
 }
 
-type encodeFunc = func(v interface{}, isErrorResponse bool) error
+type encodeMsgFunc = func(msg *jsonrpcMessage, isError bool) error
+
+type encodeBatchFunc = func(msgs []*jsonrpcMessage, isError bool) error
 
 type decodeFunc = func(v interface{}) error
 
 // NewFuncCodec creates a codec which uses the given functions to read and write. If conn
 // implements ConnRemoteAddr, log messages will use it to include the remote address of
 // the connection.
-func NewFuncCodec(conn deadlineCloser, encode encodeFunc, decode decodeFunc) ServerCodec {
+func NewFuncCodec(conn deadlineCloser, encodeMsg encodeMsgFunc, encodeBatch encodeBatchFunc, decode decodeFunc) ServerCodec {
 	codec := &jsonCodec{
-		closeCh: make(chan interface{}),
-		encode:  encode,
-		decode:  decode,
-		conn:    conn,
+		closeCh:     make(chan interface{}),
+		encodeMsg:   encodeMsg,
+		encodeBatch: encodeBatch,
+		decode:      decode,
+		conn:        conn,
 	}
 	if ra, ok := conn.(ConnRemoteAddr); ok {
 		codec.remote = ra.RemoteAddr()
@@ -211,14 +231,62 @@ func NewFuncCodec(conn deadlineCloser, encode encodeFunc, decode decodeFunc) Ser
 // NewCodec creates a codec on the given connection. If conn implements ConnRemoteAddr, log
 // messages will use it to include the remote address of the connection.
 func NewCodec(conn Conn) ServerCodec {
-	enc := json.NewEncoder(conn)
 	dec := json.NewDecoder(conn)
 	dec.UseNumber()
-
-	encode := func(v interface{}, isErrorResponse bool) error {
-		return enc.Encode(v)
+	var buf []byte
+	encodeMsg := func(msg *jsonrpcMessage, isError bool) error {
+		buf = appendMessage(buf[:0], msg)
+		buf = append(buf, '\n')
+		_, err := conn.Write(buf)
+		return err
 	}
-	return NewFuncCodec(conn, encode, dec.Decode)
+	encodeBatch := func(msgs []*jsonrpcMessage, isError bool) error {
+		buf = appendBatch(buf[:0], msgs)
+		buf = append(buf, '\n')
+		_, err := conn.Write(buf)
+		return err
+	}
+	return NewFuncCodec(conn, encodeMsg, encodeBatch, dec.Decode)
+}
+
+// appendMessage appends the JSON-RPC encoding of msg to buf.
+func appendMessage(buf []byte, msg *jsonrpcMessage) []byte {
+	buf = append(buf, `{"jsonrpc":"2.0"`...)
+	if msg.ID != nil {
+		buf = append(buf, `,"id":`...)
+		buf = append(buf, msg.ID...)
+	}
+	if msg.Method != "" {
+		buf = append(buf, `,"method":`...)
+		buf = jsonw.AppendQuotedString(buf, msg.Method)
+	}
+	if msg.Params != nil {
+		buf = append(buf, `,"params":`...)
+		buf = append(buf, msg.Params...)
+	}
+	if msg.Error != nil {
+		buf = append(buf, `,"error":`...)
+		buf = append(buf, msg.Error...)
+	}
+	if msg.Result != nil {
+		buf = append(buf, `,"result":`...)
+		buf = append(buf, msg.Result...)
+	}
+	buf = append(buf, '}')
+	return buf
+}
+
+// appendBatch appends the JSON-RPC encoding of a message batch to buf.
+func appendBatch(buf []byte, msgs []*jsonrpcMessage) []byte {
+	buf = append(buf, '[')
+	for i, msg := range msgs {
+		if i > 0 {
+			buf = append(buf, ',')
+		}
+		buf = appendMessage(buf, msg)
+	}
+	buf = append(buf, ']')
+	return buf
 }
 
 func (c *jsonCodec) peerInfo() PeerInfo {
@@ -248,7 +316,7 @@ func (c *jsonCodec) readBatch() (messages []*jsonrpcMessage, batch bool, err err
 	return messages, batch, nil
 }
 
-func (c *jsonCodec) writeJSON(ctx context.Context, v interface{}, isErrorResponse bool) error {
+func (c *jsonCodec) writeJSON(ctx context.Context, msg *jsonrpcMessage, isError bool) error {
 	c.encMu.Lock()
 	defer c.encMu.Unlock()
 
@@ -257,7 +325,19 @@ func (c *jsonCodec) writeJSON(ctx context.Context, v interface{}, isErrorRespons
 		deadline = time.Now().Add(defaultWriteTimeout)
 	}
 	c.conn.SetWriteDeadline(deadline)
-	return c.encode(v, isErrorResponse)
+	return c.encodeMsg(msg, isError)
+}
+
+func (c *jsonCodec) writeJSONBatch(ctx context.Context, msgs []*jsonrpcMessage, isError bool) error {
+	c.encMu.Lock()
+	defer c.encMu.Unlock()
+
+	deadline, ok := ctx.Deadline()
+	if !ok {
+		deadline = time.Now().Add(defaultWriteTimeout)
+	}
+	c.conn.SetWriteDeadline(deadline)
+	return c.encodeBatch(msgs, isError)
 }
 
 func (c *jsonCodec) close() {

rpc/server_test.go

@@ -208,6 +208,48 @@ func TestServerBatchResponseSizeLimit(t *testing.T) {
 	}
 }
 
+// TestServerBatchResponseSizeLimit_errorResponses verifies that error responses
+// are counted toward BatchResponseMaxSize.
+func TestServerBatchResponseSizeLimit_errorResponses(t *testing.T) {
+	t.Parallel()
+
+	server := newTestServer()
+	defer server.Stop()
+	// Each error response for test_returnError is ~58 bytes of JSON in the Error field.
+	// Set limit to 100 so 1 response fits (58 bytes) but the 2nd (116 bytes) exceeds it.
+	server.SetBatchLimits(100, 100)
+	var (
+		batch  []BatchElem
+		client = DialInProc(server)
+	)
+	for i := 0; i < 5; i++ {
+		batch = append(batch, BatchElem{
+			Method: "test_returnError",
+			Result: new(int),
+		})
+	}
+	if err := client.BatchCall(batch); err != nil {
+		t.Fatal("error sending batch:", err)
+	}
+	for i := range batch {
+		re, ok := batch[i].Error.(Error)
+		if !ok {
+			t.Fatalf("batch elem %d has wrong error type: %v", i, batch[i].Error)
+		}
+		if i < 2 {
+			// First two: elem 0 fits under limit, elem 1 pushes over but is already processed.
+			if re.ErrorCode() != 444 {
+				t.Errorf("batch elem %d wrong error code, have %d want 444", i, re.ErrorCode())
+			}
+		} else {
+			// Remaining should be the response-too-large error.
+			if re.ErrorCode() != errcodeResponseTooLarge {
+				t.Errorf("batch elem %d wrong error code, have %d want %d", i, re.ErrorCode(), errcodeResponseTooLarge)
+			}
+		}
+	}
+}
+
 func TestServerWebsocketReadLimit(t *testing.T) {
 	t.Parallel()
 

rpc/subscription.go

@@ -171,13 +171,17 @@ func (n *Notifier) activate() error {
 }
 
 func (n *Notifier) send(sub *Subscription, data any) error {
-	msg := jsonrpcSubscriptionNotification{
+	params, err := json.Marshal(subscriptionResultEnc{
+		ID:     string(sub.ID),
+		Result: data,
+	})
+	if err != nil {
+		return err
+	}
+	msg := jsonrpcMessage{
 		Version: vsn,
 		Method:  n.namespace + notificationMethodSuffix,
-		Params: subscriptionResultEnc{
-			ID:     string(sub.ID),
-			Result: data,
-		},
+		Params:  params,
 	}
 	return n.h.conn.writeJSON(context.Background(), &msg, false)
 }

rpc/subscription_test.go

@@ -220,7 +220,7 @@ func readAndValidateMessage(in *json.Decoder) (*subConfirmation, *subscriptionRe
 	case msg.isResponse():
 		var c subConfirmation
 		if msg.Error != nil {
-			return nil, nil, msg.Error
+			return nil, nil, msg.decodeError()
 		} else if err := json.Unmarshal(msg.Result, &c.subid); err != nil {
 			return nil, nil, fmt.Errorf("invalid response: %v", err)
 		} else {
@@ -233,12 +233,21 @@ func readAndValidateMessage(in *json.Decoder) (*subConfirmation, *subscriptionRe
 }
 
 type mockConn struct {
-	enc *json.Encoder
+	w io.Writer
 }
 
-// writeJSON writes a message to the connection.
-func (c *mockConn) writeJSON(ctx context.Context, msg interface{}, isError bool) error {
-	return c.enc.Encode(msg)
+func (c *mockConn) writeJSON(ctx context.Context, msg *jsonrpcMessage, isError bool) error {
+	buf := appendMessage(nil, msg)
+	buf = append(buf, '\n')
+	_, err := c.w.Write(buf)
+	return err
+}
+
+func (c *mockConn) writeJSONBatch(ctx context.Context, msgs []*jsonrpcMessage, isError bool) error {
+	buf := appendBatch(nil, msgs)
+	buf = append(buf, '\n')
+	_, err := c.w.Write(buf)
+	return err
 }
 
 // closed returns a channel which is closed when the connection is closed.
@@ -251,7 +260,7 @@ func (c *mockConn) remoteAddr() string { return "" }
 func BenchmarkNotify(b *testing.B) {
 	id := ID("test")
 	notifier := &Notifier{
-		h:         &handler{conn: &mockConn{json.NewEncoder(io.Discard)}},
+		h:         &handler{conn: &mockConn{io.Discard}},
 		sub:       &Subscription{ID: id},
 		activated: true,
 	}
@@ -271,7 +280,7 @@ func TestNotify(t *testing.T) {
 	out := new(bytes.Buffer)
 	id := ID("test")
 	notifier := &Notifier{
-		h:         &handler{conn: &mockConn{json.NewEncoder(out)}},
+		h:         &handler{conn: &mockConn{out}},
 		sub:       &Subscription{ID: id},
 		activated: true,
 	}

rpc/types.go

@@ -51,9 +51,10 @@ type ServerCodec interface {
 // jsonWriter can write JSON messages to its underlying connection.
 // Implementations must be safe for concurrent use.
 type jsonWriter interface {
-	// writeJSON writes a message to the connection.
-	writeJSON(ctx context.Context, msg interface{}, isError bool) error
-
+	// writeJSON writes a single JSON-RPC message to the connection.
+	writeJSON(ctx context.Context, msg *jsonrpcMessage, isError bool) error
+	// writeJSONBatch writes a batch of JSON-RPC messages to the connection.
+	writeJSONBatch(ctx context.Context, msgs []*jsonrpcMessage, isError bool) error
 	// Closed returns a channel which is closed when the connection is closed.
 	closed() <-chan interface{}
 	// RemoteAddr returns the peer address of the connection.

rpc/websocket.go

@@ -293,11 +293,17 @@ type websocketCodec struct {
 
 func newWebsocketCodec(conn *websocket.Conn, host string, req http.Header, readLimit int64) ServerCodec {
 	conn.SetReadLimit(readLimit)
-	encode := func(v interface{}, isErrorResponse bool) error {
-		return conn.WriteJSON(v)
+	var buf []byte
+	encodeMsg := func(msg *jsonrpcMessage, isError bool) error {
+		buf = appendMessage(buf[:0], msg)
+		return conn.WriteMessage(websocket.TextMessage, buf)
+	}
+	encodeBatch := func(msgs []*jsonrpcMessage, isError bool) error {
+		buf = appendBatch(buf[:0], msgs)
+		return conn.WriteMessage(websocket.TextMessage, buf)
 	}
 	wc := &websocketCodec{
-		jsonCodec:    NewFuncCodec(conn, encode, conn.ReadJSON).(*jsonCodec),
+		jsonCodec:    NewFuncCodec(conn, encodeMsg, encodeBatch, conn.ReadJSON).(*jsonCodec),
 		conn:         conn,
 		pingReset:    make(chan struct{}, 1),
 		pongReceived: make(chan struct{}),
@@ -342,8 +348,15 @@ func (wc *websocketCodec) peerInfo() PeerInfo {
 	return wc.info
 }
 
-func (wc *websocketCodec) writeJSON(ctx context.Context, v interface{}, isError bool) error {
-	err := wc.jsonCodec.writeJSON(ctx, v, isError)
+func (wc *websocketCodec) writeJSON(ctx context.Context, msg *jsonrpcMessage, isError bool) error {
+	return wc.writeAndResetPing(wc.jsonCodec.writeJSON(ctx, msg, isError))
+}
+
+func (wc *websocketCodec) writeJSONBatch(ctx context.Context, msgs []*jsonrpcMessage, isError bool) error {
+	return wc.writeAndResetPing(wc.jsonCodec.writeJSONBatch(ctx, msgs, isError))
+}
+
+func (wc *websocketCodec) writeAndResetPing(err error) error {
 	if err == nil {
 		// Notify pingLoop to delay the next idle ping.
 		select {