Refactor transition post genesis #311

gballet · 2023-11-24T16:44:07Z

The base branch is able to perform the transition in one block, but it fails if there are more than one block. The problem stems from the fact ended and started are global variables, and that during block creation, several attempts are made at creating a block with a different tx set each time. So the state transition pointers were incorrect. To make things worse, once the block has been created, it's passed to the CL, and the latter passes it back to the EL for insertion, at which points the block is re-executed.

This PR implements a cleaner approach in which there is only one "active" set of conversion pointers at a time, but once the block has been created, these sets of pointers (called TransitionState) are saved on a per-root basis. This way, if (for whatever reason) multiple blocks are built on top of this one, the pointers are always going to be the same.

When building on top of the parent block, the state database copies the final TransitionState of its parent, so that it can updates the pointers as it goes through the transition. When it's done, and the final, translated root root is calculated, the current TransitionState is saved into the state database list of transition states, referenced by the final root.

TODO

use a LRU list instead of an ever-growing map
check with several nodes
so far there needs to be a TransitionState for every block past the fork. It would be better if this system could be jettisoned after the conversion.
fix cycle in tests
check verkle at genesis - validate that fix: a branch that can verge at genesis or post genesis #314 works with spindevnets and merge it.
check this works in kurtosis
fix post-genesis transition verkle branch fails to join after transition #316 before merging

gballet

Left a few explanatory comments.

gballet · 2023-11-24T16:44:31Z

consensus/beacon/consensus.go

@@ -25,6 +25,7 @@ import (
 	"github.com/ethereum/go-ethereum/consensus"
 	"github.com/ethereum/go-ethereum/consensus/misc/eip1559"
 	"github.com/ethereum/go-ethereum/consensus/misc/eip4844"
+	"github.com/ethereum/go-ethereum/core"


this causes a cycle in tests

gballet · 2023-11-24T16:45:08Z

consensus/beacon/consensus.go

@@ -368,6 +369,12 @@ func (beacon *Beacon) Finalize(chain consensus.ChainHeaderReader, header *types.
 	state.Witness().TouchAddressOnWriteAndComputeGas(header.Coinbase[:], uint256.Int{}, utils.NonceLeafKey)
 	state.Witness().TouchAddressOnWriteAndComputeGas(header.Coinbase[:], uint256.Int{}, utils.CodeKeccakLeafKey)
 	state.Witness().TouchAddressOnWriteAndComputeGas(header.Coinbase[:], uint256.Int{}, utils.CodeSizeLeafKey)
+
+	if chain.Config().IsPrague(header.Number, header.Time) {
+		fmt.Println("at block", header.Number, "performing transition?", state.Database().InTransition())


Suggested change

fmt.Println("at block", header.Number, "performing transition?", state.Database().InTransition())

gballet · 2023-11-24T16:46:30Z

consensus/beacon/consensus.go

+	if chain.Config().IsPrague(header.Number, header.Time) {
+		fmt.Println("at block", header.Number, "performing transition?", state.Database().InTransition())
+		parent := chain.GetHeaderByHash(header.ParentHash)
+		core.OverlayVerkleTransition(state, parent.Root)


this PR does the same thing as #309, but it was easier to reason about when called in a single location. It should still be possible to move it to e.g. Prepare but that's definitely enough work for a single PR.

gballet · 2023-11-24T17:15:28Z

consensus/beacon/consensus.go

@@ -405,6 +413,7 @@ func (beacon *Beacon) FinalizeAndAssemble(chain consensus.ChainHeaderReader, hea
 			return nil, fmt.Errorf("nil parent header for block %d", header.Number)
 		}

+		state.Database().LoadTransitionState(parent.Root)


When producing the proof, the pre-tree is re-opened. Unfortunately, the OpenTrie function needs the pointers, who have to be reloaded. This is a bit tacky and could be avoided if the pre tree was opened before, e.g. in Prepare.

gballet · 2023-11-24T17:18:44Z

consensus/beacon/consensus.go

+			// This should only happen for the first block,
+			// so the previous tree is a merkle tree. Logically,
+			// the "previous" verkle tree is an empty tree.
+			okpre = true
+			vtrpre = trie.NewVerkleTrie(verkle.New(), state.Database().TrieDB(), utils.NewPointCache(), false)
+			post := state.GetTrie().(*trie.TransitionTrie)
+			vtrpost = post.Overlay()
+			okpost = true


the change caused this panic to trigger, because OpenTrie needs the parameters to decide if it's a verkle or merkle tree. But I don't think it was correct to begin with, because for the first block, the "pre" tree is technically a MPT. But if we want to prove against something, that something we prove against is an empty verkle tree. So this is what this code does: it returns an empty verkle tree so that the proof can be built against it.

For some reason I haven't determined yet, the first proof is still empty. That is an investigation for later though.

gballet · 2023-11-24T17:28:14Z

core/state/database.go

+	if ts.CurrentAccountAddress != nil {
+		ret.CurrentAccountAddress = &common.Address{}
+		copy(ret.CurrentAccountAddress[:], ts.CurrentAccountAddress[:])
+	}


super important: don't just copy the pointer to a unique address.

gballet · 2023-11-24T17:28:45Z

core/state/database.go

-	LastMerkleRoot common.Hash // root hash of the read-only base tree
+	LastMerkleRoot         common.Hash // root hash of the read-only base tree
+	CurrentTransitionState *TransitionState
+	TransitionStatePerRoot map[common.Hash]*TransitionState


Suggested change

TransitionStatePerRoot map[common.Hash]*TransitionState

TransitionStatePerRoot map[common.Hash]*TransitionState // List of per-root conversion state pointers

gballet · 2023-11-24T17:30:45Z

core/state/database.go

+	ts, ok := db.TransitionStatePerRoot[root]
+	if !ok || ts == nil {
+		// Start with a fresh state
+		ts = &TransitionState{ended: db.triedb.IsVerkle()}


this is where we handle the "verkle at genesis case" since if that happens, then the loaded transition state will be nil and so it is initialized.

Note that this hasn't been tested yet, but should be before everything is merged back into kaustinen-post-shapella.

gballet · 2023-11-24T17:31:05Z

core/state_processor.go

-	parent := p.bc.GetHeaderByHash(header.ParentHash)
-	if err := OverlayVerkleTransition(statedb, parent.Root); err != nil {
-		return nil, nil, 0, fmt.Errorf("error performing verkle overlay transition: %w", err)
-	}


moved the transition to Finalize.

gballet · 2023-11-24T17:31:24Z

miner/worker.go

-	if w.chain.Config().IsPrague(header.Number, header.Time) {
-		core.OverlayVerkleTransition(state, parent.Root)
-	}


moved the conversion to Finalize.

gballet · 2023-11-30T13:45:54Z

merging into parent because the devops infrastructure needs smaller branch names 😅

* rewrite per-block conversion pointer management * remove unused method * fix: a branch that can verge at genesis or post genesis (#314) * fix: import cycle in conversion refactor (#315) * fix shadowfork panic in OpenStorageTrie

quell linter issue support Transition post tree in conversion Refactor transition post genesis (#311) * rewrite per-block conversion pointer management * remove unused method * fix: a branch that can verge at genesis or post genesis (#314) * fix: import cycle in conversion refactor (#315) * fix shadowfork panic in OpenStorageTrie fix OpenStorageTrie: return an error instead of panicking if the account tree not a verkle tree (#321) add a switch to force proof in blocks (#322) * add a switch to force proof in blocks * activate switch * fix switch type add switch to override the stride of the overlay conversion (#323) * add switch to override the stride of the overlay conversion * set a default stride of 10k add a few traces for relaunch add missing step in the chain of setting the override for overlay stride fix: save and load transition state for block processing (#324) * fix: save and load transition state for block processing * log whether the tree is verkle in LoadTransitionState * fix: ensure the transition is marked as started in insertChain * dump saved address * fix nil pointer panic * remove stacktrace that is no longer useful * fix a panic * fix build * check: copy current account address BEFORE it's saved * mandatory panic fix * Remove debug fmt.Println * more cleanup + comments fix: ensure StorageProcessed is properly recovered add traces for gas pool overflow cmd/{geth, utils}: add a command to clear verkle costs (#326) * cmd/{geth, utils}: add a command to clear verkle costs fix: boolean issue fix: load finalization state in FinalizeAndAssemble (#340) * Conversion and TransitionTrie fixes (#346) * fixes Signed-off-by: Ignacio Hagopian <jsign.uy@gmail.com> * remove old comment Signed-off-by: Ignacio Hagopian <jsign.uy@gmail.com> --------- Signed-off-by: Ignacio Hagopian <jsign.uy@gmail.com> * trace cleanup --------- Signed-off-by: Ignacio Hagopian <jsign.uy@gmail.com> Co-authored-by: Ignacio Hagopian <jsign.uy@gmail.com> fix: check for nil override in NewBlockChain (#357) fix rebase issue fix a few rebase issues add debug traces fix: assume that having to create a new transaction state mean it hasn't happened. This is a workaround, because it will always be false when restarting a node that has started the transition but not necessarily completed it. add logs to debug shadowfork persist conversion state to db and use an LRU cache for active transition states (#375) * persist conversion state to db * fix: don't recreate LRU when writing state * opt: only write state to db if not already present in LRU * fix: rlp can't encode TransitionState * fix: use gob because binary.Write does not support structs 🤦‍♂️ * fix: nil pointer panic * add logs to debug shadowfork * no such thing as not enough traces * ditto * fix stupid bug * add a comment for readability * add more traces * Lock the state transition during conversion (#384) * heavy handed approach: lock the state transition during conversion * also lock transition state loading/unloading * reduce logs verbosity * add conversion test to workflow (#386) * add conversion test to workflow * mandatory -f switch fix in rm * forgot & at the end of the geth command * remove incorrect kill * add debug traces * add an overlay stride * fix typo * Apply suggestions from code review eth: add debug_conversionStatus RPC call (#392) * eth: add debug_conversionStatus RPC call * add debug trace4s * Apply suggestions from code review * export started/ended fields fix post-verge sync (#404) * fix post-verge sync * review: fix truncated comment

gballet added 2 commits November 24, 2023 17:19

rewrite per-block conversion pointer management

ff855a8

remove unused method

0cf1dfb

gballet commented Nov 24, 2023

View reviewed changes

gballet mentioned this pull request Nov 28, 2023

fix: a branch that can verge at genesis or post genesis #314

Merged

gballet added 3 commits November 30, 2023 12:37

fix: a branch that can verge at genesis or post genesis (#314)

13b84f3

fix: import cycle in conversion refactor (#315)

0da3cd3

fix shadowfork panic in OpenStorageTrie

4ee2726

gballet merged commit 1d80ebd into transition-post-genesis Nov 30, 2023

gballet mentioned this pull request Apr 29, 2024

post-genesis transition #424

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Refactor transition post genesis #311

Refactor transition post genesis #311

gballet commented Nov 24, 2023 •

edited

Loading

gballet left a comment

gballet Nov 24, 2023

gballet Nov 24, 2023

gballet Nov 24, 2023

gballet Nov 24, 2023

gballet Nov 24, 2023

gballet Nov 24, 2023

gballet Nov 24, 2023

gballet Nov 24, 2023

gballet Nov 24, 2023

gballet Nov 24, 2023

gballet commented Nov 30, 2023

	TransitionStatePerRoot map[common.Hash]*TransitionState
	TransitionStatePerRoot map[common.Hash]*TransitionState // List of per-root conversion state pointers

Refactor transition post genesis #311

Refactor transition post genesis #311

Conversation

gballet commented Nov 24, 2023 • edited Loading

TODO

gballet left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

gballet commented Nov 30, 2023

gballet commented Nov 24, 2023 •

edited

Loading