eth/tracers/native: add erc7562 tracer #31006

shahafn · 2025-01-08T14:50:57Z

Resolves #30546

rjl493456442 · 2025-01-09T09:31:07Z

Probably relevant with #30546 , @s1na please check it out.

MariusVanDerWijden

I'm not opposed the idea of adding a 4337 style native tracer, but this one needs a bunch of refactors imo

MariusVanDerWijden · 2025-01-20T15:59:45Z

eth/tracers/native/erc7562.go

@@ -0,0 +1,550 @@
+// Copyright 2021 The go-ethereum Authors


Suggested change

// Copyright 2021 The go-ethereum Authors

// Copyright 2025 The go-ethereum Authors

MariusVanDerWijden · 2025-01-20T16:21:37Z

eth/tracers/native/erc7562.go

+}
+
+type callFrameWithOpcodes struct {
+	Type         vm.OpCode       `json:"-"`


WDYT about reusing some code from the call tracer?

diff --git a/eth/tracers/native/erc7562.go b/eth/tracers/native/erc7562.go index cd1a48975f..7f80c7d168 100644 --- a/eth/tracers/native/erc7562.go +++ b/eth/tracers/native/erc7562.go @@ -48,20 +48,7 @@ type contractSizeWithOpcode struct { } type callFrameWithOpcodes struct { - Type vm.OpCode `json:"-"` - From common.Address `json:"from"` - Gas uint64 `json:"gas"` - GasUsed uint64 `json:"gasUsed"` - To *common.Address `json:"to,omitempty" rlp:"optional"` - Input []byte `json:"input" rlp:"optional"` - Output []byte `json:"output,omitempty" rlp:"optional"` - Error string `json:"error,omitempty" rlp:"optional"` - RevertReason string `json:"revertReason,omitempty"` - Logs []callLog `json:"logs,omitempty" rlp:"optional"` - // Placed at end on purpose. The RLP will be decoded to 0 instead of - // nil if there are non-empty elements after in the struct. - Value *big.Int `json:"value,omitempty" rlp:"optional"` - revertedSnapshot bool + callFrame AccessedSlots accessedSlots `json:"accessedSlots"` ExtCodeAccessInfo []common.Address `json:"extCodeAccessInfo"` @@ -228,12 +215,14 @@ func (t *erc7562Tracer) OnEnter(depth int, typ byte, from common.Address, to com toCopy := to call := callFrameWithOpcodes{ - Type: vm.OpCode(typ), - From: from, - To: &toCopy, - Input: common.CopyBytes(input), - Gas: gas, - Value: value, + callFrame: callFrame{ + Type: vm.OpCode(typ), + From: from, + To: &toCopy, + Input: common.CopyBytes(input), + Gas: gas, + Value: value, + }, AccessedSlots: accessedSlots{ Reads: map[string][]string{}, Writes: map[string]uint64{}, ```

MariusVanDerWijden · 2025-01-20T16:24:25Z

eth/tracers/native/erc7562.go

+	"github.com/holiman/uint256"
+)
+
+//go:generate go run github.com/fjl/gencodec -type callFrameWithOpcodes -field-override callFrameWithOpcodesMarshaling -out gen_callframewithopcodes_json.go


Suggested change

//go:generate go run github.com/fjl/gencodec -type callFrameWithOpcodes -field-override callFrameWithOpcodesMarshaling -out gen_callframewithopcodes_json.go

//go:generate go run github.com/fjl/gencodec -type callFrameWithOpcodes -field-override callFrameMarshaling -out gen_callframewithopcodes_json.go

go generate doesn't work with that, and it seems that embedded structs aren't supported in gencodec. So my solution was to flatten out the struct instead. Wdyt?

MariusVanDerWijden · 2025-01-20T16:25:59Z

eth/tracers/native/erc7562.go

+}
+
+// catchPanic handles panic recovery and logs the panic and stack trace.
+func catchPanic() {


I really don't like this pattern. Your code should be correct, you should not rely on recovering. Rather fail fast and fix quickly

Right, I'd like to second this. Did you need this for debug purposes?

Panics here should not crash the node. API internal has recovery mechanism from panics. If you managed to crash the node then that's an issue we should look at.

I needed that for debugging purposes, since the errors on the hooks were silently ignored by geth it seems. I'm removing it.

MariusVanDerWijden · 2025-01-20T16:26:59Z

eth/tracers/native/erc7562.go

+	return config
+}
+
+func newErc7562TracerObject(ctx *tracers.Context, cfg json.RawMessage) (*erc7562Tracer, error) {


Suggested change

func newErc7562TracerObject(ctx *tracers.Context, cfg json.RawMessage) (*erc7562Tracer, error) {

func newErc7562TracerObject(cfg json.RawMessage) (*erc7562Tracer, error) {

ctx is unused here, no need to pass it then

tracers.DefaultDirectory.Register("erc7562Tracer", newErc7562Tracer, false) expects a certain signature for the tracer that accepts that argument. That's also the case in the original callTracer

MariusVanDerWijden · 2025-01-20T16:28:29Z

eth/tracers/native/erc7562.go

+	t.callstackWithOpcodes = append(t.callstackWithOpcodes, call)
+}
+
+func (t *erc7562Tracer) captureEnd(output []byte, gasUsed uint64, err error, reverted bool) {


Suggested change

func (t *erc7562Tracer) captureEnd(output []byte, gasUsed uint64, err error, reverted bool) {

func (t *erc7562Tracer) captureEnd(output []byte, err error, reverted bool) {

gas used is unused

MariusVanDerWijden · 2025-01-20T16:32:18Z

eth/tracers/native/erc7562.go

+}
+
+func incrementCount[K comparable](m map[K]uint64, k K) {
+	m[k] = m[k] + 1


I'm sorry what?

MariusVanDerWijden · 2025-01-20T16:40:20Z

eth/tracers/native/erc7562.go

+	for i := 0; i < t.config.StackTopItemsSize && i < stackSize; i++ {
+		stackTopItems = append(stackTopItems, *peepStack(scope.StackData(), i))
+	}
+	opcodeWithStack = &opcodeWithPartialStack{


I would switch on the opcode here instead of calling into multiple methods that might or might not do anything

s1na · 2025-01-21T16:45:09Z

It would be nice to have some spec-compliance tests.

s1na · 2025-01-21T16:51:39Z

eth/tracers/native/erc7562.go

+	ignoredOpcodes       map[vm.OpCode]struct{}
+	callstackWithOpcodes []callFrameWithOpcodes
+	lastOpWithStack      *opcodeWithPartialStack
+	Keccak               map[string]struct{} `json:"keccak"`


I think this field can be un-exported and the json tag is not needed. This struct is not directly serialized.

- Removing unused code - Adding test cases

s1na · 2025-02-21T15:58:03Z

@shahafn I see that you have added some test fixtures. I think you forgot to commit the test runner. Please push that too.

s1na · 2025-02-21T16:05:43Z

eth/tracers/native/erc7562.go

+	AccessedSlots     accessedSlots                              `json:"accessedSlots"`
+	ExtCodeAccessInfo []common.Address                           `json:"extCodeAccessInfo"`
+	UsedOpcodes       map[vm.OpCode]uint64                       `json:"usedOpcodes"`
+	ContractSize      map[common.Address]*contractSizeWithOpcode `json:"contractSize"`


Can you please add a short comment on these "new" fields what they should collect?

s1na · 2025-02-21T16:06:07Z

eth/tracers/native/erc7562.go

+	Output            hexutil.Bytes
+	AccessedSlots     accessedSlots                              `json:"accessedSlots"`
+	ExtCodeAccessInfo []common.Address                           `json:"extCodeAccessInfo"`
+	UsedOpcodes       map[vm.OpCode]uint64                       `json:"usedOpcodes"`


Probably also use hexutil.Uint64 here for consistency?

s1na · 2025-02-21T16:10:08Z

eth/tracers/native/erc7562.go

+	UsedOpcodes       map[vm.OpCode]uint64                       `json:"usedOpcodes"`
+	ContractSize      map[common.Address]*contractSizeWithOpcode `json:"contractSize"`
+	OutOfGas          bool                                       `json:"outOfGas"`
+	Calls             []callFrameWithOpcodes                     `json:"calls,omitempty"`


In the Marshaling struct you only need to specify the fields which you wish to override the type for. So in this case Calls is not necessary here, Same with OutOfGas, ContractSize, ExtCodeAccessInfo and AccessedSlots.

The reason for using this double-struct approach is so that in the tracer logic we can simply use normal Go types (like uint64) and when calling json.Marshal it will "automatically" understand how to encode them because we provide hints here that encoding should use these other specialized types (e.g. hexutil.Uint64).

s1na · 2025-02-21T16:12:41Z

eth/tracers/native/erc7562.go

+type accessedSlots struct {
+	Reads           map[string][]string `json:"reads"`
+	Writes          map[string]uint64   `json:"writes"`
+	TransientReads  map[string]uint64   `json:"transientReads"`


We have to agree on the output format. I think it's not great that some fields are decimal numbers in JSON and some are hex-encoded.

Fixing marshaling

Comparing unmarshaled objects and sorting keccak array to prevent cosmetic diffs

lightclient · 2025-05-05T20:55:55Z

What do we want to do about this? On one hand it seems like a good thing for geth to offer, on another it is a complex tracer that we are not the experts in yet will need to maintain in perpetuity.

shahafn · 2025-05-07T16:22:48Z

Hey @lightclient
We're planning on maintaining the tracer ourselves if that helps!

s1na · 2025-05-18T19:00:07Z

I have pushed the changes I had in mind here: https://github.com/s1na/go-ethereum/tree/erc7562-tracer. I can't directly push to your branch, it seems the organization should provide the permission. You can alternatively integrate from my branch directly. Otherwise it looks good to me.

* refactor keccak preimages * simplify tests * rm depth field * more interrup handling * hex opcode in output * Respect ignored opcodes * minor readability change * common.Hash for storage access fields

s1na

LGTM

eth/tracers/internal/tracetest/erc7562_tracer_test.go

s1na

Second time's the charm

This PR introduces a new native tracer for AA bundlers. Bundlers participating in the alternative mempool will need to validate userops. This tracer will return sufficient information for them to decide whether griefing is possible. Resolves ethereum#30546 --------- Co-authored-by: Sina M <[email protected]>

Adding erc7562 tracer

c94a39c

shahafn requested a review from s1na as a code owner January 8, 2025 14:50

shahafn added 2 commits January 8, 2025 18:12

Remove unused function, sort imports

6c913fa

Fixing uncaught error

d0755a6

rjl493456442 assigned s1na Jan 9, 2025

MariusVanDerWijden reviewed Jan 20, 2025

View reviewed changes

s1na reviewed Jan 21, 2025

View reviewed changes

PR review fixes

9d5a502

- Removing unused code - Adding test cases

s1na requested changes Feb 21, 2025

View reviewed changes

shahafn added 3 commits March 24, 2025 00:30

Adding erc7562_tracer_test.go

d7258a9

Fixing marshaling

Fixing test

29dcafe

Ignoring key orders in objects

1785745

Comparing unmarshaled objects and sorting keccak array to prevent cosmetic diffs

lightclient added the status:triage label May 5, 2025

Updates to the ERC7562 tracer (#57)

fdd4eb8

* refactor keccak preimages * simplify tests * rm depth field * more interrup handling * hex opcode in output * Respect ignored opcodes * minor readability change * common.Hash for storage access fields

s1na removed the status:triage label May 19, 2025

s1na previously approved these changes May 19, 2025

View reviewed changes

shahafn and others added 2 commits May 19, 2025 16:37

Merge branch 'master' into erc7562-tracer

6cf0997

fix blob schedule in test files (#58)

464a1cb

shahafn dismissed s1na’s stale review via 464a1cb May 19, 2025 15:28

shahafn commented May 19, 2025

View reviewed changes

eth/tracers/internal/tracetest/erc7562_tracer_test.go Outdated Show resolved Hide resolved

shahafn and others added 2 commits May 19, 2025 19:36

Removing unused field

b377e10

fix linter (#59)

448e310

s1na approved these changes May 20, 2025

View reviewed changes

s1na changed the title ~~Adding erc7562 tracer~~ eth/tracers/native: add erc7562 tracer May 20, 2025

s1na added this to the 1.15.12 milestone May 20, 2025

s1na merged commit 62aa6b2 into ethereum:master May 20, 2025
3 of 4 checks passed

	// Copyright 2021 The go-ethereum Authors
	// Copyright 2025 The go-ethereum Authors

	//go:generate go run github.com/fjl/gencodec -type callFrameWithOpcodes -field-override callFrameWithOpcodesMarshaling -out gen_callframewithopcodes_json.go
	//go:generate go run github.com/fjl/gencodec -type callFrameWithOpcodes -field-override callFrameMarshaling -out gen_callframewithopcodes_json.go

	func newErc7562TracerObject(ctx tracers.Context, cfg json.RawMessage) (erc7562Tracer, error) {
	func newErc7562TracerObject(cfg json.RawMessage) (*erc7562Tracer, error) {

	func (t *erc7562Tracer) captureEnd(output []byte, gasUsed uint64, err error, reverted bool) {
	func (t *erc7562Tracer) captureEnd(output []byte, err error, reverted bool) {

eth/tracers/native: add erc7562 tracer #31006

eth/tracers/native: add erc7562 tracer #31006

Conversation

shahafn commented Jan 8, 2025 • edited by s1na Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

rjl493456442 commented Jan 9, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

MariusVanDerWijden left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

s1na commented Jan 21, 2025

Uh oh!

s1na Jan 21, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

s1na commented Feb 21, 2025

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

lightclient commented May 5, 2025

Uh oh!

shahafn commented May 7, 2025

Uh oh!

s1na commented May 18, 2025

Uh oh!

s1na left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

s1na left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

shahafn commented Jan 8, 2025 •

edited by s1na

Loading

rjl493456442 commented Jan 9, 2025 •

edited

Loading

s1na Jan 21, 2025 •

edited

Loading