Metachain notarizez based on equivalent proofs #6513

sstanculeanu · 2024-10-02T16:03:04Z

Reasoning behind the pull request

metachain should notarize based on equivalent proofs

Proposed changes

update processors

Testing procedure

with feat branch

Pre-requisites

Based on the Contributing Guidelines the PR author and the reviewers must check the following requirements are met:

was the PR targeted to the correct branch?
if this is a larger feature that probably needs more than one PR, is there a feat branch created?
if this is a feat branch merging, do all satellite projects have a proper tag inside go.mod?

AdoAdoAdo · 2024-10-04T11:03:05Z

process/block/baseProcess.go

@@ -639,7 +642,7 @@ func (bp *baseProcessor) sortHeaderHashesForCurrentBlockByNonce(usedInBlock bool

 	bp.hdrsForCurrBlock.mutHdrsForBlock.RLock()
 	for metaBlockHash, headerInfo := range bp.hdrsForCurrBlock.hdrHashAndInfo {
-		if headerInfo.usedInBlock != usedInBlock {
+		if headerInfo.usedInBlock != usedInBlock || !headerInfo.hasProof {


I think that if a header is used in block but has no valid proof, then this is an error case
Could you add the condition separately and log an error for that?
in case we end up in this case we need to find out how it happened and fix.
When picking up a header to notarize it, and adding it to usedInBlock then it should have already been checked that there is a valid proof for it.

usedInBlock field is set to true also when shard headers are requested.. it is basically set in advance to true, even though the header is missing.. so maybe the proof would be received later.. not sure if this is possible though

AdoAdoAdo · 2024-10-04T14:40:05Z

process/block/metablock.go

@@ -1081,8 +1089,18 @@ func (mp *metaProcessor) createAndProcessCrossMiniBlocksDstMe(
 			continue
 		}

+		shouldCheckProof := mp.enableEpochsHandler.IsFlagEnabledInEpoch(common.EquivalentMessagesFlag, currShardHdr.GetEpoch())


the computeLongestChainFromLastNotarized would return I think the chain of headers without the last one (which is considered the confirmation for the last returned one).

Was computeLongestChainFromLastNotarized updated for equivalent proofs?

updated the blockProcessor as well

AdoAdoAdo · 2024-10-04T14:56:45Z

process/block/metablock.go

+	isBlockAfterEquivalentMessagesFlag := !check.IfNil(lastMetaBlock) &&
+		mp.enableEpochsHandler.IsFlagEnabledInEpoch(common.EquivalentMessagesFlag, lastMetaBlock.GetEpoch())
+	if isBlockAfterEquivalentMessagesFlag {
+		// for the first block we need to update both the state of the previous one and for current


and for all future blocks, updateState should be done only for current, not lastMetaBlock.

I would use a different variable e.g finalMetaBlock and not overwrite the lastMetaBlock (that should still point to the last committed meta block not the one currently being committed).

should work as described, updated the variable name for clarity

AdoAdoAdo · 2024-10-04T15:09:39Z

process/block/metablock.go

@@ -1838,6 +1867,20 @@ func (mp *metaProcessor) checkShardHeadersFinality(
 			continue
 		}

+		if mp.enableEpochsHandler.IsFlagEnabledInEpoch(common.EquivalentMessagesFlag, lastVerifiedHdr.GetEpoch()) {


I don't think this loop is correct, as it does not get to iterate over all shard headers and verify.
it returns either a nil or an error in the first iteration.

true, moved this if into the next for loop and switched the return nil into a break

AdoAdoAdo · 2024-10-07T13:30:47Z

process/block/metablock.go

+			hasProof := mp.proofsPool.HasProof(shardHeader.GetShardID(), shardHeaderHash)
+			hdrInfoForHash.hasProof = hasProof
+			if hasProof {
+				mp.hdrsForCurrBlock.missingHdrsProofs--


missingHdrsProofs should be decreased if that proof was missing previously and still missing at this step only

e.g you receive a shard hdr which is not part of the metablock (an older block maybe), so here you would end up decreasing the missing headers proofs.

updated the condition, it was not working as expected

AdoAdoAdo · 2024-10-07T14:18:47Z

process/block/metablock.go

@@ -1994,13 +2059,24 @@ func (mp *metaProcessor) computeExistingAndRequestMissingShardHeaders(metaBlock
 		if hdr.GetNonce() > mp.hdrsForCurrBlock.highestHdrNonce[shardData.ShardID] {
 			mp.hdrsForCurrBlock.highestHdrNonce[shardData.ShardID] = hdr.GetNonce()
 		}
+
+		if mp.shouldConsiderProofsForNotarization(hdr) {
+			notarizedShardHdrsBasedOnProofs++


instead of incrementing the missing headers proofs, maybe use a map for the missing header proofs, with key the header hash

then you could correctly check which proofs are still missing.

AdoAdoAdo · 2024-10-07T14:25:52Z

process/block/metablock.go

+			mp.enableEpochsHandler.IsFlagEnabledInEpoch(common.EquivalentMessagesFlag, headerInfo.hdr.GetEpoch())
+		hasMissingProof := isBlockAfterEquivalentMessagesFlag && !headerInfo.hasProof
+		if hasMissingProof {
+			continue


this is an error if a header was used in block but has no proof, should not continue but return an error.
If we get such errors, then we need to fix in the place where the headers are notarized, as there should be the check that the notarized headers have the proper proof.

Please check that the shardInfo (createShardInfo) contains the proof for the shard header itself, along with the proof for the previous block.
This means that the shardInfo would change as well to include 2 proofs.

updated to return error

updated also to use both proofs

AdoAdoAdo · 2024-10-07T15:37:13Z

process/block/shardblock.go

-				core.MetachainShardId,
-				sp.metaBlockFinality,
-			)
+			if !sp.shouldConsiderProofsForNotarization(metaBlock) {


on the else, you should at least check that there is a valid proof for this metablock, otherwise add it to the missing proofs.

AdoAdoAdo · 2024-10-07T15:38:18Z

process/block/metablock.go

 		if mp.hdrsForCurrBlock.missingHdrs == 0 {
-			mp.hdrsForCurrBlock.missingFinalityAttestingHdrs = mp.requestMissingFinalityAttestingShardHeaders()
+			if !mp.shouldConsiderProofsForNotarization(shardHeader) {


on the else, you should check that there is a valid proof for this header, otherwise add it to the missing proofs.

AdoAdoAdo · 2024-10-07T15:41:29Z

process/block/shardblock.go

+			hasProofForShardHdr := sp.proofsPool.HasProof(core.MetachainShardId, metaBlockHashes[i])
+			sp.hdrsForCurrBlock.hdrHashAndInfo[string(metaBlockHashes[i])].hasProof = hasProofForShardHdr
+			if !hasProofForShardHdr {
+				sp.hdrsForCurrBlock.missingHdrsProofs++


maybe add the header hash in a map for missing proofs instead?

moved the logic to be based on hasProof field, which is part of a map too

ssd04 · 2024-10-08T11:30:30Z

process/block/hdrForBlock.go

@@ -9,6 +9,7 @@ import (
 type hdrForBlock struct {
 	missingHdrs                  uint32
 	missingFinalityAttestingHdrs uint32
+	missingHdrsProofs            uint32


i might help to rename here to indicate that it's num missing

ssd04 · 2024-10-08T11:33:05Z

process/block/metablock.go

 		if len(currShardHdr.GetMiniBlockHeadersWithDst(mp.shardCoordinator.SelfId())) == 0 {
-			mp.hdrsForCurrBlock.hdrHashAndInfo[string(orderedHdrsHashes[i])] = &hdrInfo{hdr: currShardHdr, usedInBlock: true}
+			mp.hdrsForCurrBlock.hdrHashAndInfo[string(orderedHdrsHashes[i])] = &hdrInfo{hdr: currShardHdr, usedInBlock: true, hasProof: true}


split on multiple lines for better visibility? same below

ssd04 · 2024-10-08T11:36:45Z

process/block/metablock.go

@@ -1081,8 +1089,18 @@ func (mp *metaProcessor) createAndProcessCrossMiniBlocksDstMe(
 			continue
 		}

+		shouldCheckProof := mp.enableEpochsHandler.IsFlagEnabledInEpoch(common.EquivalentMessagesFlag, currShardHdr.GetEpoch())
+		hasProofForHdr := mp.proofsPool.HasProof(currShardHdr.GetShardID(), orderedHdrsHashes[i])


i think here we should check if HasProof only if shouldCheckProof is true

correct, optimization.. implemented it

ssd04 · 2024-10-08T12:03:17Z

process/block/baseProcess.go

adapt also sortHeadersForCurrentBlockByNonce to have hasProof check?

ssd04 · 2024-10-08T13:07:12Z

process/block/baseProcess.go

@@ -639,7 +642,7 @@ func (bp *baseProcessor) sortHeaderHashesForCurrentBlockByNonce(usedInBlock bool

 	bp.hdrsForCurrBlock.mutHdrsForBlock.RLock()
 	for metaBlockHash, headerInfo := range bp.hdrsForCurrBlock.hdrHashAndInfo {
-		if headerInfo.usedInBlock != usedInBlock {
+		if headerInfo.usedInBlock != usedInBlock || !headerInfo.hasProof {


check for enabled flag here?

ssd04 · 2024-10-08T13:21:57Z

process/block/metablock.go

@@ -1838,6 +1867,20 @@ func (mp *metaProcessor) checkShardHeadersFinality(
 			continue
 		}

+		if mp.enableEpochsHandler.IsFlagEnabledInEpoch(common.EquivalentMessagesFlag, lastVerifiedHdr.GetEpoch()) {


does it fits better to check proof in checkShardHeadersValidity which is called just before instead of here?

I think it should stay here, as here is the place where we check for finality

ssd04 · 2024-10-08T13:44:18Z

process/block/shardblock.go

+
+		headerHash := sp.hasher.Compute(string(marshalledHeader))
+		if !sp.proofsPool.HasProof(header.GetShardID(), headerHash) {
+			return process.ErrHeaderNotFinal


return a composed error here to indicate also that the proof is missing?

ssd04 · 2024-10-08T13:45:37Z

process/block/shardblock.go

 		createAndProcessInfo.currMetaHdrHash = orderedMetaBlocksHashes[i]
 		if len(createAndProcessInfo.currMetaHdr.GetMiniBlockHeadersWithDst(sp.shardCoordinator.SelfId())) == 0 {
-			sp.hdrsForCurrBlock.hdrHashAndInfo[string(createAndProcessInfo.currMetaHdrHash)] = &hdrInfo{hdr: createAndProcessInfo.currMetaHdr, usedInBlock: true}
+			sp.hdrsForCurrBlock.hdrHashAndInfo[string(createAndProcessInfo.currMetaHdrHash)] = &hdrInfo{hdr: createAndProcessInfo.currMetaHdr, usedInBlock: true, hasProof: true}


suggestion to expand on multiple lines

…rsx/mx-chain-go into adapt_notarization # Conflicts: # process/block/baseProcess.go

…otarization

…rsx/mx-chain-go into adapt_notarization # Conflicts: # integrationTests/multiShard/endOfEpoch/startInEpoch/startInEpoch_test.go # integrationTests/multiShard/hardFork/hardFork_test.go # process/block/shardblock.go

…more with the new flow

ssd04 · 2024-11-14T15:14:36Z

process/block/metablock.go

+	// check proofs for shard data
+	for _, shardData := range header.ShardInfo {


i think here we should do this only if equivalent messages flag is activated

ssd04 · 2024-11-14T15:15:03Z

process/block/metablock.go

+			// if there is an entry for the missing proof, it means that proofsPool did not have it while scanning shardData
+			// thus header epoch was not available at that time
+			incompleteProof, hasMissingProof := mp.hdrsForCurrBlock.missingProofs[string(shardHeaderHash)]


i think here also

ssd04 · 2024-11-14T15:15:45Z

process/block/metablock.go

 	}

-	if mp.hdrsForCurrBlock.missingHdrs == 0 {
+	shouldRequestMissingFinalityAttestingShardHeaders := notarizedShardHdrsBasedOnProofs != len(metaBlock.ShardInfo)


check this only if equivalent flag enabled?

not needed, here notarizedShardHdrsBasedOnProofs would be 0, thus the second condition will always be true

ssd04 · 2024-11-14T15:16:03Z

process/block/shardblock.go

+	// check proofs for shard data
+	for _, metaBlockHash := range header.GetMetaBlockHashes() {


check if proofs activated?

ssd04 · 2024-11-14T15:16:20Z

process/block/shardblock.go

 	}

-	if sp.hdrsForCurrBlock.missingHdrs == 0 {
+	shouldRequestMissingFinalityAttestingMetaHeaders := notarizedMetaHdrsBasedOnProofs != len(metaBlockHashes)


add this only if flag activated

this should be ok too

now metachain notarizez based on equivalent proofs

1eae0e9

sstanculeanu added type:feature New feature or request area:processing impact:medium interested:protocol complexity:medium labels Oct 2, 2024

sstanculeanu self-assigned this Oct 2, 2024

sstanculeanu marked this pull request as ready for review October 3, 2024 06:53

sstanculeanu added 2 commits October 3, 2024 12:30

Merge branch 'feat/equivalent-messages' into adapt_notarization

4197d2b

Merge branch 'feat/equivalent-messages' into adapt_notarization

71187d1

AdoAdoAdo reviewed Oct 7, 2024

View reviewed changes

ssd04 reviewed Oct 8, 2024

View reviewed changes

sstanculeanu and others added 13 commits October 10, 2024 10:45

Merge branch 'feat/equivalent-messages' of https://github.com/multive…

367d1d1

…rsx/mx-chain-go into adapt_notarization # Conflicts: # process/block/baseProcess.go

Merge remote-tracking branch 'origin/adapt_notarization' into adapt_n…

3df3761

…otarization

fixes after review

8236823

fixed more tests

d96462a

fix after merge

2e1b145

Merge branch 'feat/equivalent-messages' into adapt_notarization

f34b5cf

fix tests

6aad53b

subround block fixes

50ba8a1

removed tests for after equivalent messages as they are not valid any…

3fb045b

…more with the new flow

fix linter

93afc4e

remove one more test

1158d33

proper proofs verification on notarization

450d335

ssd04 requested changes Nov 14, 2024

View reviewed changes

fixes after review

644d8be

		// check proofs for shard data
		for _, shardData := range header.ShardInfo {

		// check proofs for shard data
		for _, metaBlockHash := range header.GetMetaBlockHashes() {

Metachain notarizez based on equivalent proofs #6513

Are you sure you want to change the base?

Metachain notarizez based on equivalent proofs #6513

Conversation

sstanculeanu commented Oct 2, 2024

Reasoning behind the pull request

Proposed changes

Testing procedure

Pre-requisites

Choose a reason for hiding this comment

sstanculeanu Oct 10, 2024 • edited Loading

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

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

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

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

sstanculeanu Oct 10, 2024 •

edited

Loading