go-ethereum/core/txpool/blobpool/cache.go
Bosul Mun d91b71fb36
core/txpool/blobpool, eth: implement sparse blobpool (#34047)
This is the implementation of EIP-8070 Sparse Blobpool. It introduces
protocol version eth/72 which relays blob transaction cells instead of
full blobs.

The blobpool now store 'incomplete' transactions, where only some of
the cells are provided. The stored cell indexes are taken from the
custody bitmap, which is provided by the consensus layer in
forkchoiceUpdatedV4. This method will be called once Glamsterdam
activates, and the default custody is full custody, so for now there
is no change in the amount of stored cells for now.

The main entities added are the BlobBuffer and BlobFetcher, which work
together to track and fetch missing cells from connected peers. The
partial transactions become available for inclusion in blocks when
they are covered by enough peers that hold all cells.

This change also introduces engine_getBlobsV4, which allows for
cell-based responses (and partial blobs with only some of the cells).
We maintain backward compatibility with getBlobsV3 which expects full
blob responses by recovering the blob from available cells. Since this
process is resource-intensive, we proactively cache the conversion so
it is ready in time for getBlobsV3 calls. This mechanism will be
removed once support for getBlobsV4 is universal across all consensus
layer implementations.

devp2p tests for eth/72 are not part of this initial change. This is
to avoid breaking test success status for execution clients that do
not have eth/72 implemented yet. The tests will be added in a
subsequent change.

---------

Co-authored-by: Felix Lange <fjl@twurst.com>
2026-07-15 14:41:04 +02:00

588 lines
16 KiB
Go

// 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 blobpool
import (
"context"
"sync"
"time"
"github.com/ethereum/go-ethereum/common"
"github.com/ethereum/go-ethereum/common/mclock"
"github.com/ethereum/go-ethereum/consensus/misc/eip1559"
"github.com/ethereum/go-ethereum/consensus/misc/eip4844"
"github.com/ethereum/go-ethereum/core/txpool"
"github.com/ethereum/go-ethereum/core/txpool/txorder"
"github.com/ethereum/go-ethereum/core/types"
"github.com/ethereum/go-ethereum/crypto/kzg4844"
"github.com/ethereum/go-ethereum/internal/telemetry"
"github.com/ethereum/go-ethereum/log"
"github.com/ethereum/go-ethereum/metrics"
"github.com/holiman/uint256"
)
const (
topKTimeout = 1 * time.Second
hasBlobsTimeout = 1 * time.Second
)
var (
// Cache tracks 3 metrics: cache hit, cache miss, and the number of blobs
// it contains. Note that cache miss includes the blobs that we are actually
// missing on the lower level (in this case, the blobpool). The amount that
// we failed to predict) can be calculated with the telemetry span
// (blobs.filled - cache.hit).
cacheHitMeter = metrics.NewRegisteredMeter("blobpool/cache/hit", nil)
cacheMissMeter = metrics.NewRegisteredMeter("blobpool/cache/miss", nil)
cacheEntriesGauge = metrics.NewRegisteredGauge("blobpool/cache/entries", nil)
)
type cachedBlob struct {
cell []kzg4844.Cell
custody types.CustodyBitmap
blob *kzg4844.Blob
commitment kzg4844.Commitment
proofs []kzg4844.Proof
version byte
}
// Cache holds the blobs that are likely to be requested by the GetBlobs engine API.
//
// Every `topKTimeout`, the cache selects the blobs of the top K most profitable
// transactions, and preloads them into the cache.
//
// For HasBlobs requests, it also causes the blobs requested by the CL to be loaded.
// (Note: the cache is not guaranteed to always hold such blobs, since the blobpool might
// drop the transaction in the window between the engine API response and the cache
// update.)
type Cache struct {
blobpool *BlobPool
clock mclock.Clock
mu sync.Mutex
entries map[common.Hash]*cachedBlob
// needCell is owned by the loop goroutine; it is only read and written
// there, so it needs no synchronization. It is flipped on via enableCellCh.
needCell bool
// channels into loop
quit chan struct{}
topkRequest chan struct{}
topkTimer mclock.Timer
hasBlobsCh chan []common.Hash // list of tx hashes that should be pinned
cellModeCh chan bool // signals the loop to switch cell mode on/offo
step func() // test hook fired after each loop iteration
cancelInflights context.CancelFunc // cancels the conversion/decode goroutines
inflight sync.WaitGroup // tracks all in-flight conversion/decode goroutines
wg sync.WaitGroup // tracks the loop goroutine
}
// NewCache creates a blob cache backed by the given blobpool.
func NewCache(p *BlobPool) *Cache {
return newCache(p, mclock.System{}, nil)
}
// newCache creates a blob cache for testing purposes.
// It allows injecting a clock and a step hook.
func newCache(p *BlobPool, clock mclock.Clock, step func()) *Cache {
c := &Cache{
entries: make(map[common.Hash]*cachedBlob),
blobpool: p,
hasBlobsCh: make(chan []common.Hash, 1),
clock: clock,
step: step,
quit: make(chan struct{}),
topkRequest: make(chan struct{}, 1),
cellModeCh: make(chan bool, 1),
}
c.wg.Add(1)
go c.loop()
return c
}
// Stop terminates the cache loop and blocks until it and any in-flight work
// have stopped.
func (c *Cache) Stop() {
close(c.quit)
c.wg.Wait()
}
// HasBlobs reports whether the blob is available (in the cache or the
// blobpool) and asks the loop to pin the ones it found.
func (c *Cache) HasBlobs(ctx context.Context, vhashes []common.Hash) []bool {
var (
missIdx []int
missVhashes []common.Hash
needPin []common.Hash // available vhashes
)
available := make([]bool, len(vhashes))
// First check cache and pass missing ones to blobpool.
c.mu.Lock()
for i, vhash := range vhashes {
if _, ok := c.entries[vhash]; ok {
available[i] = true
needPin = append(needPin, vhash)
} else {
missIdx = append(missIdx, i)
missVhashes = append(missVhashes, vhash)
}
}
c.mu.Unlock()
if len(missVhashes) > 0 {
pooled := c.blobpool.availableBlobs(missVhashes)
// Merge two results
for j, ok := range pooled {
if ok {
available[missIdx[j]] = true
needPin = append(needPin, missVhashes[j])
}
}
}
select {
case c.hasBlobsCh <- needPin:
// Note that we also send the ones we already have in cache,
// since it can be dropped from the cache before this signal is processed.
return available
case <-c.quit:
return nil
}
}
// GetBlobs returns the blobs and proofs for the given versioned hashes, serving
// them from the cache when possible and falling back to the blobpool for misses.
// Responses are placed in the order given in the request, using null for any
// missing blob.
//
// For instance, if the request is [A_versioned_hash, B_versioned_hash,
// C_versioned_hash] and blobpool has data for blobs A and C, but doesn't have
// data for B, the response MUST be [A, null, C].
//
// This is a utility method for the engine API, enabling consensus clients to
// retrieve blobs from the pools directly instead of the network.
//
// The version argument specifies the type of proofs to return, either the
// blob proofs (version 0) or the cell proofs (version 1). Proofs conversion is
// CPU intensive and prohibited explicitly.
func (c *Cache) GetBlobs(ctx context.Context, vhashes []common.Hash, version byte) (_ []*kzg4844.Blob, _ []kzg4844.Commitment, _ [][]kzg4844.Proof, err error) {
_, span, spanEnd := telemetry.StartSpan(ctx, "blobpool.GetBlobs")
defer spanEnd(&err)
var (
blobs = make([]*kzg4844.Blob, len(vhashes))
commitments = make([]kzg4844.Commitment, len(vhashes))
proofs = make([][]kzg4844.Proof, len(vhashes))
indices = make(map[common.Hash][]int)
misses []common.Hash
cacheHits int
cacheMiss int
)
for i, h := range vhashes {
indices[h] = append(indices[h], i)
}
c.mu.Lock()
for vhash, idxs := range indices {
n := len(idxs)
cached, ok := c.entries[vhash]
if !ok || cached.version != version || cached.blob == nil {
cacheMiss += n
misses = append(misses, vhash)
continue
}
cacheHits += n
for _, index := range idxs {
blobs[index] = cached.blob
commitments[index] = cached.commitment
proofs[index] = cached.proofs
}
}
c.mu.Unlock()
if len(misses) > 0 {
mb, mc, mp, err := c.blobpool.getBlobs(misses, version)
if err != nil {
return nil, nil, nil, err
}
for j, vhash := range misses {
if mb[j] == nil {
continue
}
for _, index := range indices[vhash] {
blobs[index] = mb[j]
commitments[index] = mc[j]
proofs[index] = mp[j]
}
}
}
cacheHitMeter.Mark(int64(cacheHits))
cacheMissMeter.Mark(int64(cacheMiss))
span.SetAttributes(
telemetry.IntAttribute("cache.hit", cacheHits),
telemetry.IntAttribute("cache.miss", cacheMiss),
)
return blobs, commitments, proofs, nil
}
// GetCells returns cells for the given versioned blob hashes, filtered
// by the requested cell indices (mask). Each entry in the result
// corresponds to one vhash. Nil entries mean the cell was not available.
// If the cell is not available in cache, it falls back to the blobpool.
func (c *Cache) GetCells(vhashes []common.Hash, mask types.CustodyBitmap) ([][]*kzg4844.Cell, [][]*kzg4844.Proof, error) {
var (
cells = make([][]*kzg4844.Cell, len(vhashes))
proofs = make([][]*kzg4844.Proof, len(vhashes))
indices = make(map[common.Hash][]int)
misses []common.Hash
)
for i, h := range vhashes {
indices[h] = append(indices[h], i)
}
requested := mask.Indices()
c.mu.Lock()
for vhash, idxs := range indices {
cached, ok := c.entries[vhash]
if !ok {
misses = append(misses, vhash)
continue
}
stored := cached.custody.Indices()
blobCells := make([]*kzg4844.Cell, len(requested))
blobProofs := make([]*kzg4844.Proof, len(requested))
for i, cellIdx := range requested {
pos := -1
for k, s := range stored {
if s == cellIdx {
pos = k
break
}
}
if pos >= 0 && pos < len(cached.cell) {
cell := cached.cell[pos]
blobCells[i] = &cell
if int(cellIdx) < len(cached.proofs) {
pf := cached.proofs[cellIdx]
blobProofs[i] = &pf
}
}
}
for _, idx := range idxs {
cells[idx] = blobCells
proofs[idx] = blobProofs
}
}
c.mu.Unlock()
if len(misses) > 0 {
mc, mp, err := c.blobpool.GetBlobCells(misses, mask)
if err != nil {
return nil, nil, err
}
for j, vhash := range misses {
for _, idx := range indices[vhash] {
cells[idx] = mc[j]
proofs[idx] = mp[j]
}
}
}
return cells, proofs, nil
}
// EnableCell allows the cache to store only cells without recovering
// blobs. This means we can also cache cells that lack enough blobs to
// recover. It signals the loop to switch to cell mode and re-select
// transactions from this wider pool.
func (c *Cache) SetCellMode(on bool) {
select {
case c.cellModeCh <- on:
case <-c.quit:
}
}
func (c *Cache) loop() {
defer c.wg.Done()
c.triggerTopK()
for {
select {
case want := <-c.hasBlobsCh:
// HasBlobs request was received.
// Update the cache once with the requested blobs, then reschedule topK.
c.update(want)
c.triggerTopKAfter(hasBlobsTimeout)
case <-c.topkRequest:
want := c.selectTopTxs()
c.update(want)
c.triggerTopKAfter(topKTimeout)
case on := <-c.cellModeCh:
// This runs when the CL signals (non-)support for cell proofs. Enable/disable
// cell mode and re-select immediately to force an update.
if c.needCell != on {
c.needCell = on
c.triggerTopK()
}
case <-c.quit:
c.cancelUpdate()
if c.topkTimer != nil {
c.topkTimer.Stop()
}
c.inflight.Wait()
return
}
if c.step != nil {
c.step()
}
}
}
// cancelUpdate stops the current update.
func (c *Cache) cancelUpdate() {
if c.cancelInflights != nil {
c.cancelInflights()
c.cancelInflights = nil
}
}
// update updates the cache to hold the wanted vhashes. It evicts entries that
// are no longer wanted and loads the missing ones from the blobpool in the
// background.
func (c *Cache) update(want []common.Hash) {
cellMode := c.needCell
wantSet := make(map[common.Hash]struct{}, len(want))
for _, vh := range want {
wantSet[vh] = struct{}{}
}
// Cancel the current updates.
c.cancelUpdate()
ctx, cancel := context.WithCancel(context.Background())
c.cancelInflights = cancel
c.mu.Lock()
var missing []common.Hash
for vh := range wantSet {
e, ok := c.entries[vh]
if ok && ((cellMode && e.cell == nil) || (!cellMode && e.blob == nil)) {
delete(c.entries, vh)
cacheEntriesGauge.Dec(1)
ok = false
}
if !ok {
missing = append(missing, vh)
}
}
for vh := range c.entries {
if _, ok := wantSet[vh]; ok {
continue
}
delete(c.entries, vh)
cacheEntriesGauge.Dec(1)
}
c.mu.Unlock()
c.inflight.Add(1)
go func() {
defer c.inflight.Done()
for _, vh := range missing {
select {
case <-ctx.Done():
return
default:
}
c.mu.Lock()
_, loaded := c.entries[vh]
c.mu.Unlock()
if loaded {
continue
}
ptx := c.blobpool.getByVhash(vh)
if ptx == nil {
continue
}
if cellMode {
c.loadCells(ptx, wantSet)
} else {
c.loadBlobs(ptx, wantSet)
}
}
}()
}
// loadCells loads the cells whose vhash is in wantSet from ptx.
func (c *Cache) loadCells(ptx *BlobTxForPool, wantSet map[common.Hash]struct{}) {
cs := ptx.CellSidecar
cellsPerBlob := cs.Custody.OneCount()
c.mu.Lock()
defer c.mu.Unlock()
for i, v := range ptx.Tx.BlobHashes() {
if _, ok := wantSet[v]; !ok {
continue
}
if _, exists := c.entries[v]; exists {
continue
}
cellStart := i * cellsPerBlob
if cellStart+cellsPerBlob > len(cs.Cells) {
continue
}
blobCells := make([]kzg4844.Cell, cellsPerBlob)
copy(blobCells, cs.Cells[cellStart:cellStart+cellsPerBlob])
var pf []kzg4844.Proof
if ps := i * kzg4844.CellProofsPerBlob; ps+kzg4844.CellProofsPerBlob <= len(cs.Proofs) {
pf = make([]kzg4844.Proof, kzg4844.CellProofsPerBlob)
copy(pf, cs.Proofs[ps:ps+kzg4844.CellProofsPerBlob])
}
c.entries[v] = &cachedBlob{
cell: blobCells,
custody: cs.Custody,
commitment: cs.Commitments[i],
proofs: pf,
version: cs.Version,
}
cacheEntriesGauge.Inc(1)
}
}
// loadBlobs loads the blobs whose vhash is in wantSet from ptx.
func (c *Cache) loadBlobs(ptx *BlobTxForPool, wantSet map[common.Hash]struct{}) {
if ptx.CellSidecar.Custody.OneCount() < kzg4844.DataPerBlob {
return
}
// blobs will be computed inside of sidecar()
sidecar, err := ptx.sidecar()
if err != nil || sidecar == nil {
return
}
c.mu.Lock()
defer c.mu.Unlock()
for i, v := range sidecar.BlobHashes() {
if _, ok := wantSet[v]; !ok {
continue
}
if _, exists := c.entries[v]; exists {
continue
}
var pf []kzg4844.Proof
switch sidecar.Version {
case types.BlobSidecarVersion0:
pf = []kzg4844.Proof{sidecar.Proofs[i]}
case types.BlobSidecarVersion1:
cellProofs, err := sidecar.CellProofsAt(i)
if err != nil {
log.Error("Failed to get cell proofs", "txhash", ptx.Tx.Hash(), "err", err)
continue
}
pf = cellProofs
}
c.entries[v] = &cachedBlob{
blob: &sidecar.Blobs[i],
commitment: sidecar.Commitments[i],
proofs: pf,
version: sidecar.Version,
}
cacheEntriesGauge.Inc(1)
}
}
// selectTopTxs returns the vhashes of the top K most profitable pending blob
// transactions, up to the active fork's maxBlobsPerBlock.
func (c *Cache) selectTopTxs() []common.Hash {
p := c.blobpool
head := p.head.Load()
if head == nil {
return nil
}
config := p.chain.Config()
baseFee := eip1559.CalcBaseFee(config, head)
filter := txpool.PendingFilter{
BlobTxs: true,
BaseFee: uint256.MustFromBig(baseFee),
PartialCells: c.needCell,
}
if head.ExcessBlobGas != nil {
filter.BlobFee = uint256.MustFromBig(eip4844.CalcBlobFee(config, head))
}
if config.IsOsaka(head.Number, head.Time) {
filter.BlobVersion = types.BlobSidecarVersion1
} else {
filter.BlobVersion = types.BlobSidecarVersion0
}
pending, _ := p.Pending(filter)
vhashesOf := p.vhashesByTx()
order := txorder.NewTransactionsByPriceAndNonce(p.signer, pending, baseFee)
// Bound the selection by the active fork's blob limit so the cache follows
// BPO changes to maxBlobsPerBlock.
target := uint(eip4844.MaxBlobsPerBlock(config, head.Time))
var (
vhashes []common.Hash
blobs uint
)
for blobs < target {
tx, _ := order.Peek()
if tx == nil {
break
}
vh, ok := vhashesOf[tx.Hash]
if ok {
vhashes = append(vhashes, vh...)
blobs += uint(len(vh))
}
order.Shift()
}
return vhashes
}
// triggerTopKAfter makes a topK selection happen after the given interval.
func (c *Cache) triggerTopKAfter(interval time.Duration) {
if c.topkTimer != nil {
c.topkTimer.Stop()
}
// drain current request to avoid triggering before the interval
select {
case <-c.topkRequest:
default:
}
c.topkTimer = c.clock.AfterFunc(interval, c.triggerTopK)
}
// triggerTopK causes another topK selection to happen.
// Note this is safe to call from anywhere, even outside of the loop goroutine.
func (c *Cache) triggerTopK() {
select {
case c.topkRequest <- struct{}{}:
default:
}
}