ICS4: Fix Sequencing logic in UpgradeTry #5419

AdityaSripal · 2023-12-14T14:36:25Z

Description

While writing quint tests against the spec I noticed a few issues surrounding our sequencing logic. This PR fixes them.

Issues:

Unexpected Behaviour: UpgradeTry persists state writes after returning errorReceipt. This will cause the TRY chain to (for example) initialize with the upgrade attempt of an out-of-date counterparty that it wants to abort
Inefficiency: In the crossing hello case, we do not fastforward the sequence. This will create an issue in the following circumstance. Suppose two sides have the exact same upgrade but different sequences, our current logic will fail to allow the upgrade to proceed. Instead, we just move the fastforwarding logic outside the crossing hello check and do it regardless so long as the upgrades are compatible.

See the following issues on spec:
cosmos/ibc#1052
cosmos/ibc#1053
cosmos/ibc#1054

This PR on the spec address the issues:
cosmos/ibc#1055

Unit tests to write:

(this is already written)
Case 1: ChainA(Sequence: 1), ChainB(Sequence 5)
ChainA.INIT
ChainB.TRY(ERROR_RECEIPT(5))
ChainA.Cancel(5)
ChainA.INIT(6)
ChainB.TRY(6)
...

Case 2: Equal sequences should work. I believe this is already written

Case 3: ChainA(Sequence: 5), ChainB(Sequence: 1)
ChainA.INIT
ChainB.TRY // fastforward to 5
...

// CROSSING HELLO CHECK

Case 4: ChainA(Sequence: 1), ChainB(Sequence 5)
ChainA.INIT
ChainB.INIT
ChainB.TRY(ErrorReceipt: 4)
ChainA.Cancel(4)
ChainA.TRY(5)
...

Case 5: ChainA(Sequence: 1), ChainB(Sequence 5)
ChainA.INIT
ChainB.INIT
ChainA.TRY(5) // fastforward
ChainB.TRY(ErrorReceipt: 4)
ChainA.Cancel => Error
ChainB.Ack => Success
...

// This test may already be written
Case 6: Crossing hello with the same sequence and upgrade should work fine.

// INCOMPATIBLE CROSSING HELLO

Case 7: ChainA(Sequence: 1, ics20-2), ChainB(Sequence: 5, ics20-3)
ChainA.INIT
ChainB.INIT
ChainA.TRY => error
ChainB.TRY(ErrorReceipt: 4)
ChainA.Cancel(4)
ChainA.TRY(5)
....

Case 8: ChainA(Sequence: 1, ics20-2), ChainB(Sequence: 1, ics20-3)
ChainA.INIT
ChainB.INIT
ChainA.TRY => error
ChainB.TRY => error
ChainA.INIT(2, ics20-3)
ChainB.TRY(2, ics20-3)
....

Commit Message / Changelog Entry

type: commit message

see the guidelines for commit messages. (view raw markdown for examples)

Before we can merge this PR, please make sure that all the following items have been
checked off. If any of the checklist items are not applicable, please leave them but
write a little note why.

Targeted PR against correct branch (see CONTRIBUTING.md).
Linked to Github issue with discussion and accepted design OR link to spec that describes this work.
Code follows the module structure standards and Go style guide.
Wrote unit and integration tests.
Updated relevant documentation (docs/) or specification (x/<module>/spec/).
Added relevant godoc comments.
Provide a commit message to be used for the changelog entry in the PR description for review.
Re-reviewed Files changed in the Github PR explorer.
Review Codecov Report in the comment section below once CI passes.

crodriguezvega · 2023-12-15T14:01:23Z

The following test should cover some of the cases mentioned:

Case 1 by TestChanUpgrade_UpgradeSucceeds_AfterCancel.
Case 2 by success of TestChanUpgradeTry.
Case 3 by success: upgrade sequence is fast forwarded to counterparty upgrade sequence of TestChanUpgradeTry.

And then case 6 by TestChanUpgradeCrossingHelloWithHistoricalProofs.

And I added tests for cases 4, 5 7, and 8.

crodriguezvega · 2023-12-15T22:18:56Z

I get "account sequence mismatch, expected 14, got 13: incorrect account sequence" in of the tests, if anybody can figure out why, that would be amazing. :)

damiannolan · 2023-12-18T12:59:38Z

The account sequence mismatch error is arising because we only increment the account sequence in the testing pkg if the tx goes through successfully.

In the test in which its failing, we call the upgrade cancel handler, and expect an error, but later try to send another msg to the chain with a stale signing sequence.

https://github.com/cosmos/ibc-go/blob/main/testing/chain.go#L373-L377

edit: we could probably wrap the above code in a defer to ensure that the account sequence is incremented when the func returns, whether there is an error present or not.

edit2: I modified the testing pkg code to use the defer and the test then passes!

crodriguezvega · 2023-12-18T20:01:28Z

Thank you for finding the problem, @damiannolan! Would you like to push the fix?

…uccess and failure tx execution

damiannolan

ACK, looks good to me, happy to move forward with this! LGTM

I can cherry-pick the commit to main for TestChain.SendMsgs() if needs be, but I think its probably fine to keep it in the feat branch and just let it get merged through there!

damiannolan · 2023-12-20T08:46:04Z

modules/core/04-channel/keeper/upgrade.go

+		// The ChanUpgradeTry message may arrive first, in this case, **IF** the upgrade fields are mutually compatible; the counterparty will simply
+		// fast forward their sequence to our own and continue the upgrade. The following ChanUpgradeCancel message will be rejected as it is below the current sequence.
+
+		return channel, upgrade, types.NewUpgradeError(upgradeSequence, errorsmod.Wrapf(


Its intensional that the channel and upgrade are returned here instead of empty structs, right?

This is the only place where an upgrade error is return from the try handler and it looks like the channel returned is used for emitting error receipt events in this case:
https://github.com/cosmos/ibc-go/blob/04-channel-upgrades/modules/core/keeper/msg_server.go#L818-L820

modules/core/04-channel/keeper/upgrade_test.go

DimitrisJim

lgtm, cov is also clean, amazing! 💯

DimitrisJim · 2023-12-20T09:09:57Z

modules/core/04-channel/keeper/upgrade.go

 	if counterpartyUpgradeSequence < channel.UpgradeSequence {
-		return channel, upgrade, types.NewUpgradeError(channel.UpgradeSequence-1, errorsmod.Wrapf(
+		// In this case, the counterparty upgrade is outdated. We want to force the counterparty


delicious documentation @AdityaSripal, very helpful

DimitrisJim · 2023-12-20T09:16:21Z

testing/chain.go

@@ -344,6 +344,14 @@ func (chain *TestChain) SendMsgs(msgs ...sdk.Msg) (*abci.ExecTxResult, error) {
 	// ensure the chain has the latest time
 	chain.Coordinator.UpdateTimeForChain(chain)

+	// increment acc sequence regardless of success or failure tx execution


great find @damiannolan 👀

…un and add more assertions

damiannolan · 2023-12-20T09:57:07Z

Added more assertions to tests and wrapped steps in suite.Run to try make the flow as clear as possible 👍🏻

fix sequencing logic and unit tests

d22d046

AdityaSripal requested review from colin-axner and damiannolan as code owners December 14, 2023 14:36

AdityaSripal marked this pull request as draft December 14, 2023 14:36

colin-axner added channel-upgradability Channel upgradability feature priority PRs that need prompt reviews labels Dec 14, 2023

Merge branch '04-channel-upgrades' into aditya/fix-sequencing

baac838

crodriguezvega self-assigned this Dec 15, 2023

Carlos Rodriguez added 2 commits December 15, 2023 14:55

added tests

699af0c

reorder comments

b46879a

Carlos Rodriguez added 3 commits December 15, 2023 16:38

add test

7eb246c

fix typo

f8aa85b

refactored a couple of tests

bae2f59

Carlos Rodriguez added 2 commits December 15, 2023 23:20

relocate comment

da7629f

Merge branch '04-channel-upgrades' into aditya/fix-sequencing

728615c

Carlos Rodriguez and others added 2 commits December 19, 2023 11:01

Merge branch '04-channel-upgrades' into aditya/fix-sequencing

a711f06

fix: defer incrementing acc sequencing such that is happens on both s…

57ab98b

…uccess and failure tx execution

damiannolan marked this pull request as ready for review December 19, 2023 10:11

damiannolan requested review from crodriguezvega, charleenfei, chatton and DimitrisJim as code owners December 19, 2023 10:11

crodriguezvega removed their assignment Dec 19, 2023

Merge branch '04-channel-upgrades' into aditya/fix-sequencing

09a50c4

damiannolan approved these changes Dec 20, 2023

View reviewed changes

chore: update version strings in comments

a5174d1

damiannolan reviewed Dec 20, 2023

View reviewed changes

modules/core/04-channel/keeper/upgrade_test.go Outdated Show resolved Hide resolved

DimitrisJim approved these changes Dec 20, 2023

View reviewed changes

damiannolan added 2 commits December 20, 2023 10:24

test: wrap test code in suite.Run and add additional assertions

39b7262

refactor: adapt crossing hellos cancellation test case to use suite.R…

aefb902

…un and add more assertions

AdityaSripal merged commit ab1f96f into 04-channel-upgrades Dec 20, 2023
56 checks passed

AdityaSripal deleted the aditya/fix-sequencing branch December 20, 2023 10:48

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

ICS4: Fix Sequencing logic in UpgradeTry #5419

ICS4: Fix Sequencing logic in UpgradeTry #5419

AdityaSripal commented Dec 14, 2023 •

edited

Loading

crodriguezvega commented Dec 15, 2023 •

edited

Loading

crodriguezvega commented Dec 15, 2023

damiannolan commented Dec 18, 2023 •

edited

Loading

crodriguezvega commented Dec 18, 2023

damiannolan left a comment

damiannolan Dec 20, 2023

DimitrisJim left a comment

DimitrisJim Dec 20, 2023

DimitrisJim Dec 20, 2023

damiannolan commented Dec 20, 2023

ICS4: Fix Sequencing logic in UpgradeTry #5419

ICS4: Fix Sequencing logic in UpgradeTry #5419

Conversation

AdityaSripal commented Dec 14, 2023 • edited Loading

Description

Commit Message / Changelog Entry

crodriguezvega commented Dec 15, 2023 • edited Loading

crodriguezvega commented Dec 15, 2023

damiannolan commented Dec 18, 2023 • edited Loading

crodriguezvega commented Dec 18, 2023

damiannolan left a comment

Choose a reason for hiding this comment

damiannolan Dec 20, 2023

Choose a reason for hiding this comment

DimitrisJim left a comment

Choose a reason for hiding this comment

DimitrisJim Dec 20, 2023

Choose a reason for hiding this comment

DimitrisJim Dec 20, 2023

Choose a reason for hiding this comment

damiannolan commented Dec 20, 2023

AdityaSripal commented Dec 14, 2023 •

edited

Loading

crodriguezvega commented Dec 15, 2023 •

edited

Loading

damiannolan commented Dec 18, 2023 •

edited

Loading