Merge pull request #11400 from filecoin-project/phi/v1250rc4

build: Calib hot fix: v1.25.0-rc4

CHANGELOG.md

@@ -1,14 +1,18 @@
 # Lotus changelog
 
-# v 1.25.0-rc3 / 2023-11-02
+# v 1.25.0-rc4 / 2023-11-08
 
-This is the second release candidate of the upcoming OPTIONAL release Lotus v1.25.0. This optional release also supports the Filecoin network version 21 upgrade, codenamed Watermelon 🍉, in addition to the numerous improvements and enhancements for node operators, ETH RPC-providers and storage providers.
+This is the fourth release candidate of the upcoming OPTIONAL release Lotus v1.25.0. This optional release also supports the Filecoin network version 21 upgrade, codenamed Watermelon 🍉, in addition to the numerous improvements and enhancements for node operators, ETH RPC-providers and storage providers.
 
-This release candidate also sets an upgrade epoch for mainnet at 3431940 2023-11-29T13:30:00Z.
+## Calibration Testnet WatermelonFix Upgrade Recovery
 
-It includes a patch for the calibration testnet to fix an issue where partitions with expired sectors had to be compacted before they could be moved. Unfortunately, this bug impacts consensus and necessitates a coordinated upgrade on the Calibration network to deploy the new code. The calibration network is scheduled to upgrade all miner actors to the new, fixed miner actor CID at epoch 1070494, which is expected to occur at 2023-11-07T13:00:00Z.
+The Calibration Testnet halted 60 epochs after the WatermelonFix upgrade, we believe the cause is the new fixed miner actor CID isn't registered in the system actor state. Fortunately, this could be fixedby winding back the time prior to the upgrade, and reperform the migration with new miner actor CID registered in the system actor state. We would like to ask all calibrationnet node operators to run the following:
+- Upgrade your nodes and miners to v1.25.0-rc4
+- Shut down your daemon and restart it (miners too if applicable)
 
-Make sure to check out the release log for [Lotus v1.24.0-rc4](https://github.com/filecoin-project/lotus/releases/tag/v1.24.0-rc4) to see the FIPs delivered in the network version 21 upgrade, v12 Builtin Actor Bundles, features and improvements.
+Please reach out to us in #fil-net-calibration-discuss if you have any questions!
+
+Make sure to check out the release log for [Lotus v1.24.0-rc5](https://github.com/filecoin-project/lotus/releases/tag/v1.24.0-rc5) to see the FIPs delivered in the network version 21 upgrade, v12 Builtin Actor Bundles, features and improvements.
 
 ## ☢️ Upgrade Warnings ☢️
 

build/version.go

@@ -37,7 +37,7 @@ func BuildTypeString() string {
 }
 
 // BuildVersion is the local build version
-const BuildVersion = "1.25.0-rc3"
+const BuildVersion = "1.25.0-rc4"
 
 func UserVersion() string {
 	if os.Getenv("LOTUS_VERSION_IGNORE_COMMIT") == "1" {

chain/consensus/filcns/upgrades.go

@@ -2063,6 +2063,19 @@ func upgradeActorsV12Fix(ctx context.Context, sm *stmgr.StateManager, cache stmg
 		return cid.Undef, xerrors.Errorf("failed to perform migration: %w", err)
 	}
 
+	systemState.BuiltinActors = newManifest.Data
+	newSystemHead, err := adtStore.Put(ctx, &systemState)
+	if err != nil {
+		return cid.Undef, xerrors.Errorf("failed to put new system state: %w", err)
+	}
+
+	systemActor.Head = newSystemHead
+	if err = actorsOut.SetActor(builtin.SystemActorAddr, systemActor); err != nil {
+		return cid.Undef, xerrors.Errorf("failed to put new system actor: %w", err)
+	}
+
+	// Sanity checking
+
 	err = actorsIn.ForEach(func(a address.Address, inActor *types.Actor) error {
 		outActor, err := actorsOut.GetActor(a)
 		if err != nil {
@@ -2081,7 +2094,7 @@ func upgradeActorsV12Fix(ctx context.Context, sm *stmgr.StateManager, cache stmg
 			return xerrors.Errorf("mismatched address for actor %s: %s != %s", a, inActor.Address, outActor.Address)
 		}
 
-		if inActor.Head != outActor.Head {
+		if inActor.Head != outActor.Head && a != builtin.SystemActorAddr {
 			return xerrors.Errorf("mismatched head for actor %s", a)
 		}
 

chain/stmgr/forks.go

@@ -178,16 +178,18 @@ func (sm *StateManager) HandleStateForks(ctx context.Context, root cid.Cid, heig
 	retCid := root
 	u := sm.stateMigrations[height]
 	if u != nil && u.upgrade != nil {
-		migCid, ok, err := u.migrationResultCache.Get(ctx, root)
-		if err == nil {
-			if ok {
-				log.Infow("CACHED migration", "height", height, "from", root, "to", migCid)
-				return migCid, nil
+		if height != build.UpgradeWatermelonFixHeight {
+			migCid, ok, err := u.migrationResultCache.Get(ctx, root)
+			if err == nil {
+				if ok {
+					log.Infow("CACHED migration", "height", height, "from", root, "to", migCid)
+					return migCid, nil
+				}
+			} else if !errors.Is(err, datastore.ErrNotFound) {
+				log.Errorw("failed to lookup previous migration result", "err", err)
+			} else {
+				log.Debug("no cached migration found, migrating from scratch")
 			}
-		} else if !errors.Is(err, datastore.ErrNotFound) {
-			log.Errorw("failed to lookup previous migration result", "err", err)
-		} else {
-			log.Debug("no cached migration found, migrating from scratch")
 		}
 
 		startTime := time.Now()
@@ -195,6 +197,7 @@ func (sm *StateManager) HandleStateForks(ctx context.Context, root cid.Cid, heig
 		// Yes, we clone the cache, even for the final upgrade epoch. Why? Reverts. We may
 		// have to migrate multiple times.
 		tmpCache := u.cache.Clone()
+		var err error
 		retCid, err = u.upgrade(ctx, sm, tmpCache, cb, root, height, ts)
 		if err != nil {
 			log.Errorw("FAILED migration", "height", height, "from", root, "error", err)

chain/vm/fvm.go

@@ -92,6 +92,15 @@ func (x *FvmExtern) VerifyConsensusFault(ctx context.Context, a, b, extra []byte
 		log.Info("invalid consensus fault: submitted blocks are the same")
 		return ret, totalGas
 	}
+
+	// workaround chain halt
+	if build.IsNearUpgrade(blockA.Height, build.UpgradeWatermelonFixHeight) {
+		return ret, totalGas
+	}
+	if build.IsNearUpgrade(blockB.Height, build.UpgradeWatermelonFixHeight) {
+		return ret, totalGas
+	}
+
 	// (1) check conditions necessary to any consensus fault
 
 	// were blocks mined by same miner?

documentation/en/cli-lotus-miner.md

@@ -7,7 +7,7 @@ USAGE:
    lotus-miner [global options] command [command options] [arguments...]
 
 VERSION:
-   1.25.0-rc3
+   1.25.0-rc4
 
 COMMANDS:
    init     Initialize a lotus miner repo

documentation/en/cli-lotus-worker.md

@@ -7,7 +7,7 @@ USAGE:
    lotus-worker [global options] command [command options] [arguments...]
 
 VERSION:
-   1.25.0-rc3
+   1.25.0-rc4
 
 COMMANDS:
    run        Start lotus worker

documentation/en/cli-lotus.md

@@ -7,7 +7,7 @@ USAGE:
    lotus [global options] command [command options] [arguments...]
 
 VERSION:
-   1.25.0-rc3
+   1.25.0-rc4
 
 COMMANDS:
    daemon   Start a lotus daemon process

miner/miner.go

@@ -325,7 +325,7 @@ minerLoop:
 					"block-time", btime, "time", build.Clock.Now(), "difference", build.Clock.Since(btime))
 			}
 
-			if os.Getenv("LOTUS_MINER_NO_SLASHFILTER") != "_yes_i_know_i_can_and_probably_will_lose_all_my_fil_and_power_" {
+			if os.Getenv("LOTUS_MINER_NO_SLASHFILTER") != "_yes_i_know_i_can_and_probably_will_lose_all_my_fil_and_power_" && !build.IsNearUpgrade(base.TipSet.Height(), build.UpgradeWatermelonFixHeight) {
 				witness, fault, err := m.sf.MinedBlock(ctx, b.Header, base.TipSet.Height()+base.NullRounds)
 				if err != nil {
 					log.Errorf("<!!> SLASH FILTER ERRORED: %s", err)

node/impl/full/sync.go

@@ -57,7 +57,7 @@ func (a *SyncAPI) SyncSubmitBlock(ctx context.Context, blk *types.BlockMsg) erro
 		return xerrors.Errorf("loading parent block: %w", err)
 	}
 
-	if a.SlashFilter != nil && os.Getenv("LOTUS_NO_SLASHFILTER") != "_yes_i_know_i_can_and_probably_will_lose_all_my_fil_and_power_" {
+	if a.SlashFilter != nil && os.Getenv("LOTUS_NO_SLASHFILTER") != "_yes_i_know_i_can_and_probably_will_lose_all_my_fil_and_power_" && !build.IsNearUpgrade(blk.Header.Height, build.UpgradeWatermelonFixHeight) {
 		witness, fault, err := a.SlashFilter.MinedBlock(ctx, blk.Header, parent.Height)
 		if err != nil {
 			log.Errorf("<!!> SLASH FILTER ERRORED: %s", err)