Fix Stuck Collator Funds (#4229)

Fixes https://github.com/paritytech/polkadot-sdk/issues/4206

In #1340 one of the storage types was changed from `Candidates` to
`CandidateList`. Since the actual key includes the hash of this value,
all of the candidates stored here are (a) "missing" and (b) unable to
unreserve their candidacy bond.

This migration kills the storage values and refunds the deposit held for
each candidate.

---------

Signed-off-by: georgepisaltu <george.pisaltu@parity.io>
Co-authored-by: georgepisaltu <52418509+georgepisaltu@users.noreply.github.com>
Co-authored-by: Oliver Tale-Yazdi <oliver.tale-yazdi@parity.io>
Co-authored-by: georgepisaltu <george.pisaltu@parity.io>

cumulus/pallets/collator-selection/Cargo.toml

@@ -27,6 +27,7 @@ sp-staking = { path = "../../../substrate/primitives/staking", default-features
 frame-support = { path = "../../../substrate/frame/support", default-features = false }
 frame-system = { path = "../../../substrate/frame/system", default-features = false }
 pallet-authorship = { path = "../../../substrate/frame/authorship", default-features = false }
+pallet-balances = { path = "../../../substrate/frame/balances", default-features = false }
 pallet-session = { path = "../../../substrate/frame/session", default-features = false }
 
 frame-benchmarking = { path = "../../../substrate/frame/benchmarking", default-features = false, optional = true }
@@ -38,7 +39,6 @@ sp-tracing = { path = "../../../substrate/primitives/tracing" }
 sp-runtime = { path = "../../../substrate/primitives/runtime" }
 pallet-timestamp = { path = "../../../substrate/frame/timestamp" }
 sp-consensus-aura = { path = "../../../substrate/primitives/consensus/aura" }
-pallet-balances = { path = "../../../substrate/frame/balances" }
 pallet-aura = { path = "../../../substrate/frame/aura" }
 
 [features]
@@ -59,6 +59,7 @@ std = [
 	"frame-system/std",
 	"log/std",
 	"pallet-authorship/std",
+	"pallet-balances/std",
 	"pallet-session/std",
 	"rand/std",
 	"scale-info/std",

cumulus/pallets/collator-selection/src/lib.rs

@@ -121,7 +121,7 @@ pub mod pallet {
 	use sp_std::vec::Vec;
 
 	/// The in-code storage version.
-	const STORAGE_VERSION: StorageVersion = StorageVersion::new(1);
+	const STORAGE_VERSION: StorageVersion = StorageVersion::new(2);
 
 	type BalanceOf<T> =
 		<<T as Config>::Currency as Currency<<T as SystemConfig>::AccountId>>::Balance;

cumulus/pallets/collator-selection/src/migration.rs

@@ -17,9 +17,107 @@
 //! A module that is responsible for migration of storage for Collator Selection.
 
 use super::*;
-use frame_support::traits::OnRuntimeUpgrade;
+use frame_support::traits::{OnRuntimeUpgrade, UncheckedOnRuntimeUpgrade};
 use log;
 
+/// Migrate to v2. Should have been part of <https://github.com/paritytech/polkadot-sdk/pull/1340>.
+pub mod v2 {
+	use super::*;
+	use frame_support::{
+		pallet_prelude::*,
+		storage_alias,
+		traits::{Currency, ReservableCurrency},
+	};
+	use sp_runtime::traits::{Saturating, Zero};
+	#[cfg(feature = "try-runtime")]
+	use sp_std::vec::Vec;
+
+	/// [`UncheckedMigrationToV2`] wrapped in a
+	/// [`VersionedMigration`](frame_support::migrations::VersionedMigration), ensuring the
+	/// migration is only performed when on-chain version is 1.
+	pub type MigrationToV2<T> = frame_support::migrations::VersionedMigration<
+		1,
+		2,
+		UncheckedMigrationToV2<T>,
+		Pallet<T>,
+		<T as frame_system::Config>::DbWeight,
+	>;
+
+	#[storage_alias]
+	pub type Candidates<T: Config> = StorageValue<
+		Pallet<T>,
+		BoundedVec<CandidateInfo<<T as frame_system::Config>::AccountId, <<T as Config>::Currency as Currency<<T as frame_system::Config>::AccountId>>::Balance>, <T as Config>::MaxCandidates>,
+		ValueQuery,
+	>;
+
+	/// Migrate to V2.
+	pub struct UncheckedMigrationToV2<T>(sp_std::marker::PhantomData<T>);
+	impl<T: Config + pallet_balances::Config> UncheckedOnRuntimeUpgrade for UncheckedMigrationToV2<T> {
+		fn on_runtime_upgrade() -> Weight {
+			let mut weight = Weight::zero();
+			let mut count: u64 = 0;
+			// candidates who exist under the old `Candidates` key
+			let candidates = Candidates::<T>::take();
+
+			// New candidates who have registered since the upgrade. Under normal circumstances,
+			// this should not exist because the migration should be applied when the upgrade
+			// happens. But in Polkadot/Kusama we messed this up, and people registered under
+			// `CandidateList` while their funds were locked in `Candidates`.
+			let new_candidate_list = CandidateList::<T>::get();
+			if new_candidate_list.len().is_zero() {
+				// The new list is empty, so this is essentially being applied correctly. We just
+				// put the candidates into the new storage item.
+				CandidateList::<T>::put(&candidates);
+				// 1 write for the new list
+				weight.saturating_accrue(T::DbWeight::get().reads_writes(0, 1));
+			} else {
+				// Oops, the runtime upgraded without the migration. There are new candidates in
+				// `CandidateList`. So, let's just refund the old ones and assume they have already
+				// started participating in the new system.
+				for candidate in candidates {
+					let err = T::Currency::unreserve(&candidate.who, candidate.deposit);
+					if err > Zero::zero() {
+						log::error!(
+							target: LOG_TARGET,
+							"{:?} balance was unable to be unreserved from {:?}",
+							err, &candidate.who,
+						);
+					}
+					count.saturating_inc();
+				}
+				weight.saturating_accrue(
+					<<T as pallet_balances::Config>::WeightInfo as pallet_balances::WeightInfo>::force_unreserve().saturating_mul(count.into()),
+				);
+			}
+
+			log::info!(
+				target: LOG_TARGET,
+				"Unreserved locked bond of {} candidates, upgraded storage to version 2",
+				count,
+			);
+
+			weight.saturating_accrue(T::DbWeight::get().reads_writes(3, 2));
+			weight
+		}
+
+		#[cfg(feature = "try-runtime")]
+		fn pre_upgrade() -> Result<Vec<u8>, sp_runtime::DispatchError> {
+			let number_of_candidates = Candidates::<T>::get().to_vec().len();
+			Ok((number_of_candidates as u32).encode())
+		}
+
+		#[cfg(feature = "try-runtime")]
+		fn post_upgrade(_number_of_candidates: Vec<u8>) -> Result<(), sp_runtime::DispatchError> {
+			let new_number_of_candidates = Candidates::<T>::get().to_vec().len();
+			assert_eq!(
+				new_number_of_candidates, 0 as usize,
+				"after migration, the candidates map should be empty"
+			);
+			Ok(())
+		}
+	}
+}
+
 /// Version 1 Migration
 /// This migration ensures that any existing `Invulnerables` storage lists are sorted.
 pub mod v1 {
@@ -90,3 +188,136 @@ pub mod v1 {
 		}
 	}
 }
+
+#[cfg(all(feature = "try-runtime", test))]
+mod tests {
+	use super::*;
+	use crate::{
+		migration::v2::Candidates,
+		mock::{new_test_ext, Balances, Test},
+	};
+	use frame_support::{
+		traits::{Currency, ReservableCurrency, StorageVersion},
+		BoundedVec,
+	};
+	use sp_runtime::traits::ConstU32;
+
+	#[test]
+	fn migrate_to_v2_with_new_candidates() {
+		new_test_ext().execute_with(|| {
+			let storage_version = StorageVersion::new(1);
+			storage_version.put::<Pallet<Test>>();
+
+			let one = 1u64;
+			let two = 2u64;
+			let three = 3u64;
+			let deposit = 10u64;
+
+			// Set balance to 100
+			Balances::make_free_balance_be(&one, 100u64);
+			Balances::make_free_balance_be(&two, 100u64);
+			Balances::make_free_balance_be(&three, 100u64);
+
+			// Reservations: 10 for the "old" candidacy and 10 for the "new"
+			Balances::reserve(&one, 10u64).unwrap(); // old
+			Balances::reserve(&two, 20u64).unwrap(); // old + new
+			Balances::reserve(&three, 10u64).unwrap(); // new
+
+			// Candidate info
+			let candidate_one = CandidateInfo { who: one, deposit };
+			let candidate_two = CandidateInfo { who: two, deposit };
+			let candidate_three = CandidateInfo { who: three, deposit };
+
+			// Storage lists
+			let bounded_candidates =
+				BoundedVec::<CandidateInfo<u64, u64>, ConstU32<20>>::try_from(vec![
+					candidate_one.clone(),
+					candidate_two.clone(),
+				])
+				.expect("it works");
+			let bounded_candidate_list =
+				BoundedVec::<CandidateInfo<u64, u64>, ConstU32<20>>::try_from(vec![
+					candidate_two.clone(),
+					candidate_three.clone(),
+				])
+				.expect("it works");
+
+			// Set storage
+			Candidates::<Test>::put(bounded_candidates);
+			CandidateList::<Test>::put(bounded_candidate_list.clone());
+
+			// Sanity check
+			assert_eq!(Balances::free_balance(one), 90);
+			assert_eq!(Balances::free_balance(two), 80);
+			assert_eq!(Balances::free_balance(three), 90);
+
+			// Run migration
+			v2::MigrationToV2::<Test>::on_runtime_upgrade();
+
+			let new_storage_version = StorageVersion::get::<Pallet<Test>>();
+			assert_eq!(new_storage_version, 2);
+
+			// 10 should have been unreserved from the old candidacy
+			assert_eq!(Balances::free_balance(one), 100);
+			assert_eq!(Balances::free_balance(two), 90);
+			assert_eq!(Balances::free_balance(three), 90);
+			// The storage item should be gone
+			assert!(Candidates::<Test>::get().is_empty());
+			// The new storage item should be preserved
+			assert_eq!(CandidateList::<Test>::get(), bounded_candidate_list);
+		});
+	}
+
+	#[test]
+	fn migrate_to_v2_without_new_candidates() {
+		new_test_ext().execute_with(|| {
+			let storage_version = StorageVersion::new(1);
+			storage_version.put::<Pallet<Test>>();
+
+			let one = 1u64;
+			let two = 2u64;
+			let deposit = 10u64;
+
+			// Set balance to 100
+			Balances::make_free_balance_be(&one, 100u64);
+			Balances::make_free_balance_be(&two, 100u64);
+
+			// Reservations
+			Balances::reserve(&one, 10u64).unwrap(); // old
+			Balances::reserve(&two, 10u64).unwrap(); // old
+
+			// Candidate info
+			let candidate_one = CandidateInfo { who: one, deposit };
+			let candidate_two = CandidateInfo { who: two, deposit };
+
+			// Storage lists
+			let bounded_candidates =
+				BoundedVec::<CandidateInfo<u64, u64>, ConstU32<20>>::try_from(vec![
+					candidate_one.clone(),
+					candidate_two.clone(),
+				])
+				.expect("it works");
+
+			// Set storage
+			Candidates::<Test>::put(bounded_candidates.clone());
+
+			// Sanity check
+			assert_eq!(Balances::free_balance(one), 90);
+			assert_eq!(Balances::free_balance(two), 90);
+
+			// Run migration
+			v2::MigrationToV2::<Test>::on_runtime_upgrade();
+
+			let new_storage_version = StorageVersion::get::<Pallet<Test>>();
+			assert_eq!(new_storage_version, 2);
+
+			// Nothing changes deposit-wise
+			assert_eq!(Balances::free_balance(one), 90);
+			assert_eq!(Balances::free_balance(two), 90);
+			// The storage item should be gone
+			assert!(Candidates::<Test>::get().is_empty());
+			// The new storage item should have the info now
+			assert_eq!(CandidateList::<Test>::get(), bounded_candidates);
+		});
+	}
+}

cumulus/parachains/runtimes/assets/asset-hub-rococo/src/lib.rs

@@ -973,10 +973,10 @@ pub type UncheckedExtrinsic =
 /// Migrations to apply on runtime upgrade.
 #[allow(deprecated)]
 pub type Migrations = (
-	pallet_collator_selection::migration::v1::MigrateToV1<Runtime>,
 	InitStorageVersions,
 	// unreleased
 	cumulus_pallet_xcmp_queue::migration::v4::MigrationToV4<Runtime>,
+	pallet_collator_selection::migration::v2::MigrationToV2<Runtime>,
 	// permanent
 	pallet_xcm::migration::MigrateToLatestXcmVersion<Runtime>,
 );

cumulus/parachains/runtimes/assets/asset-hub-westend/src/lib.rs

@@ -963,7 +963,7 @@ pub type Migrations = (
 	// v9420
 	pallet_nfts::migration::v1::MigrateToV1<Runtime>,
 	// unreleased
-	pallet_collator_selection::migration::v1::MigrateToV1<Runtime>,
+	pallet_collator_selection::migration::v2::MigrationToV2<Runtime>,
 	// unreleased
 	pallet_multisig::migrations::v1::MigrateToV1<Runtime>,
 	// unreleased

cumulus/parachains/runtimes/bridge-hubs/bridge-hub-rococo/src/lib.rs

@@ -139,7 +139,7 @@ pub type UncheckedExtrinsic =
 
 /// Migrations to apply on runtime upgrade.
 pub type Migrations = (
-	pallet_collator_selection::migration::v1::MigrateToV1<Runtime>,
+	pallet_collator_selection::migration::v2::MigrationToV2<Runtime>,
 	pallet_multisig::migrations::v1::MigrateToV1<Runtime>,
 	InitStorageVersions,
 	cumulus_pallet_xcmp_queue::migration::v4::MigrationToV4<Runtime>,

cumulus/parachains/runtimes/bridge-hubs/bridge-hub-westend/src/lib.rs

@@ -118,7 +118,7 @@ pub type UncheckedExtrinsic =
 
 /// Migrations to apply on runtime upgrade.
 pub type Migrations = (
-	pallet_collator_selection::migration::v1::MigrateToV1<Runtime>,
+	pallet_collator_selection::migration::v2::MigrationToV2<Runtime>,
 	pallet_multisig::migrations::v1::MigrateToV1<Runtime>,
 	InitStorageVersions,
 	// unreleased

cumulus/parachains/runtimes/collectives/collectives-westend/src/lib.rs

@@ -722,7 +722,7 @@ pub type UncheckedExtrinsic =
 /// `OnRuntimeUpgrade`. Included migrations must be idempotent.
 type Migrations = (
 	// unreleased
-	pallet_collator_selection::migration::v1::MigrateToV1<Runtime>,
+	pallet_collator_selection::migration::v2::MigrationToV2<Runtime>,
 	// unreleased
 	cumulus_pallet_xcmp_queue::migration::v4::MigrationToV4<Runtime>,
 	// permanent

cumulus/parachains/runtimes/contracts/contracts-rococo/src/lib.rs

@@ -98,6 +98,8 @@ pub type UncheckedExtrinsic =
 
 /// Migrations to apply on runtime upgrade.
 pub type Migrations = (
+	pallet_collator_selection::migration::v1::MigrateToV1<Runtime>,
+	pallet_collator_selection::migration::v2::MigrationToV2<Runtime>,
 	cumulus_pallet_parachain_system::migration::Migration<Runtime>,
 	cumulus_pallet_xcmp_queue::migration::v2::MigrationToV2<Runtime>,
 	cumulus_pallet_xcmp_queue::migration::v3::MigrationToV3<Runtime>,

cumulus/parachains/runtimes/coretime/coretime-rococo/src/lib.rs

@@ -108,6 +108,7 @@ pub type UncheckedExtrinsic =
 
 /// Migrations to apply on runtime upgrade.
 pub type Migrations = (
+	pallet_collator_selection::migration::v2::MigrationToV2<Runtime>,
 	cumulus_pallet_xcmp_queue::migration::v4::MigrationToV4<Runtime>,
 	pallet_broker::migration::MigrateV0ToV1<Runtime>,
 	// permanent

cumulus/parachains/runtimes/coretime/coretime-westend/src/lib.rs

@@ -108,6 +108,7 @@ pub type UncheckedExtrinsic =
 
 /// Migrations to apply on runtime upgrade.
 pub type Migrations = (
+	pallet_collator_selection::migration::v2::MigrationToV2<Runtime>,
 	cumulus_pallet_xcmp_queue::migration::v4::MigrationToV4<Runtime>,
 	pallet_broker::migration::MigrateV0ToV1<Runtime>,
 	// permanent

cumulus/parachains/runtimes/people/people-rococo/src/lib.rs

@@ -102,6 +102,7 @@ pub type UncheckedExtrinsic =
 
 /// Migrations to apply on runtime upgrade.
 pub type Migrations = (
+	pallet_collator_selection::migration::v2::MigrationToV2<Runtime>,
 	// permanent
 	pallet_xcm::migration::MigrateToLatestXcmVersion<Runtime>,
 );

cumulus/parachains/runtimes/people/people-westend/src/lib.rs

@@ -102,6 +102,7 @@ pub type UncheckedExtrinsic =
 
 /// Migrations to apply on runtime upgrade.
 pub type Migrations = (
+	pallet_collator_selection::migration::v2::MigrationToV2<Runtime>,
 	// permanent
 	pallet_xcm::migration::MigrateToLatestXcmVersion<Runtime>,
 );

prdoc/pr_4229.prdoc

@@ -0,0 +1,10 @@
+title: "Fix Stuck Collator Funds"
+
+doc:
+  - audience: Runtime Dev
+    description: |
+      Fixes stuck collator funds by providing a migration that should have been in PR 1340.
+
+crates:
+  - name: pallet-collator-selection
+    bump: patch