Skip to content
  1. Feb 28, 2024
    • Liam Aharon's avatar
      Runtime Upgrade ref docs and Single Block Migration example pallet (#1554) · 12ce4f7d
      Liam Aharon authored 
      Closes https://github.com/paritytech/polkadot-sdk-docs/issues/55

- Changes 'current storage version' terminology to less ambiguous
'in-code storage version' (suggestion by @ggwpez

)
- Adds a new example pallet `pallet-example-single-block-migrations`
- Adds a new reference doc to replace
https://docs.substrate.io/maintain/runtime-upgrades/ (temporarily living
in the pallet while we wait for developer hub PR to merge)
- Adds documentation for the `storage_alias` macro
- Improves `trait Hooks` docs 
- Improves `trait GetStorageVersion` docs
- Update the suggested patterns for using `VersionedMigration`, so that
version unchecked migrations are never exported
- Prevents accidental usage of version unchecked migrations in runtimes

https://github.com/paritytech/substrate/pull/14421#discussion_r1255467895
- Unversioned migration code is kept inside `mod version_unchecked`,
versioned code is kept in `pub mod versioned`
- It is necessary to use modules to limit visibility because the inner
migration must be `pub`. See
https://github.com/rust-lang/rust/issues/30905 and

https://internals.rust-lang.org/t/lang-team-minutes-private-in-public-rules/4504/40
for more.

### todo

- [x] move to reference docs to proper place within sdk-docs (now that
https://github.com/paritytech/polkadot-sdk/pull/2102 is merged)
- [x] prdoc

---------

Co-authored-by: default avatarKian Paimani <[email protected]>
Co-authored-by: default avatarJuan <[email protected]>
Co-authored-by: default avatarOliver Tale-Yazdi <[email protected]>
Co-authored-by: command-bot <>
Co-authored-by: default avatargupnik <[email protected]>
      12ce4f7d
  3. Feb 14, 2024
  5. Jan 31, 2024
    • Branislav Kontur's avatar
      [frame] `#[pallet::composite_enum]` improved variant count handling + removed... · bb8ddc46
      Branislav Kontur authored 
      [frame] `#[pallet::composite_enum]` improved variant count handling + removed `pallet_balances`'s `MaxHolds` config (#2657)

I started this investigation/issue based on @liamaharon

 question
[here](https://github.com/paritytech/polkadot-sdk/pull/1801#discussion_r1410452499).

## Problem

The `pallet_balances` integrity test should correctly detect that the
runtime has correct distinct `HoldReasons` variant count. I assume the
same situation exists for RuntimeFreezeReason.

It is not a critical problem, if we set `MaxHolds` with a sufficiently
large value, everything should be ok. However, in this case, the
integrity_test check becomes less useful.

**Situation for "any" runtime:**
- `HoldReason` enums from different pallets:
```rust
        /// from pallet_nis
        #[pallet::composite_enum]
	pub enum HoldReason {
		NftReceipt,
	}

        /// from pallet_preimage
        #[pallet::composite_enum]
	pub enum HoldReason {
		Preimage,
	}

        // from pallet_state-trie-migration
        #[pallet::composite_enum]
	pub enum HoldReason {
		SlashForContinueMigrate,
		SlashForMigrateCustomTop,
		SlashForMigrateCustomChild,
	}
```

- generated `RuntimeHoldReason` enum looks like:
```rust
pub enum RuntimeHoldReason {

    #[codec(index = 32u8)]
    Preimage(pallet_preimage::HoldReason),

    #[codec(index = 38u8)]
    Nis(pallet_nis::HoldReason),

    #[codec(index = 42u8)]
    StateTrieMigration(pallet_state_trie_migration::HoldReason),
}
```

- composite enum `RuntimeHoldReason` variant count is detected as `3`
- we set `type MaxHolds = ConstU32<3>`
- `pallet_balances::integrity_test` is ok with `3`(at least 3)

However, the real problem can occur in a live runtime where some
functionality might stop working. This is due to a total of 5 distinct
hold reasons (for pallets with multi-instance support, it is even more),
and not all of them can be used because of an incorrect `MaxHolds`,
which is deemed acceptable according to the `integrity_test`:
  ```
  // pseudo-code - if we try to call all of these:

T::Currency::hold(&pallet_nis::HoldReason::NftReceipt.into(),
&nft_owner, deposit)?;
T::Currency::hold(&pallet_preimage::HoldReason::Preimage.into(),
&nft_owner, deposit)?;

T::Currency::hold(&pallet_state_trie_migration::HoldReason::SlashForContinueMigrate.into(),
&nft_owner, deposit)?;

  // With `type MaxHolds = ConstU32<3>` these two will fail

T::Currency::hold(&pallet_state_trie_migration::HoldReason::SlashForMigrateCustomTop.into(),
&nft_owner, deposit)?;

T::Currency::hold(&pallet_state_trie_migration::HoldReason::SlashForMigrateCustomChild.into(),
&nft_owner, deposit)?;
  ```  


## Solutions

A macro `#[pallet::*]` expansion is extended of `VariantCount`
implementation for the `#[pallet::composite_enum]` enum type. This
expansion generates the `VariantCount` implementation for pallets'
`HoldReason`, `FreezeReason`, `LockId`, and `SlashReason`. Enum variants
must be plain enum values without fields to ensure a deterministic
count.

The composite runtime enum, `RuntimeHoldReason` and
`RuntimeFreezeReason`, now sets `VariantCount::VARIANT_COUNT` as the sum
of pallets' enum `VariantCount::VARIANT_COUNT`:
```rust
#[frame_support::pallet(dev_mode)]
mod module_single_instance {

	#[pallet::composite_enum]
	pub enum HoldReason {
		ModuleSingleInstanceReason1,
		ModuleSingleInstanceReason2,
	}
...
}

#[frame_support::pallet(dev_mode)]
mod module_multi_instance {

	#[pallet::composite_enum]
	pub enum HoldReason<I: 'static = ()> {
		ModuleMultiInstanceReason1,
		ModuleMultiInstanceReason2,
		ModuleMultiInstanceReason3,
	}
...
}


impl self::sp_api_hidden_includes_construct_runtime::hidden_include::traits::VariantCount
    for RuntimeHoldReason
{
    const VARIANT_COUNT: u32 = 0
        + module_single_instance::HoldReason::VARIANT_COUNT
        + module_multi_instance::HoldReason::<module_multi_instance::Instance1>::VARIANT_COUNT
        + module_multi_instance::HoldReason::<module_multi_instance::Instance2>::VARIANT_COUNT
        + module_multi_instance::HoldReason::<module_multi_instance::Instance3>::VARIANT_COUNT;
}
```

In addition, `MaxHolds` is removed (as suggested
[here](https://github.com/paritytech/polkadot-sdk/pull/2657#discussion_r1443324573))
from `pallet_balances`, and its `Holds` are now bounded to
`RuntimeHoldReason::VARIANT_COUNT`. Therefore, there is no need to let
the runtime specify `MaxHolds`.


## For reviewers

Relevant changes can be found here:
- `substrate/frame/support/procedural/src/lib.rs` 
-  `substrate/frame/support/procedural/src/pallet/parse/composite.rs`
-  `substrate/frame/support/procedural/src/pallet/expand/composite.rs`
-
`substrate/frame/support/procedural/src/construct_runtime/expand/composite_helper.rs`
-
`substrate/frame/support/procedural/src/construct_runtime/expand/hold_reason.rs`
-
`substrate/frame/support/procedural/src/construct_runtime/expand/freeze_reason.rs`
- `substrate/frame/support/src/traits/misc.rs`

And the rest of the files is just about removed `MaxHolds` from
`pallet_balances`

## Next steps

Do the same for `MaxFreezes`
https://github.com/paritytech/polkadot-sdk/issues/2997.

---------

Co-authored-by: command-bot <>
Co-authored-by: default avatarBastian Köcher <[email protected]>
Co-authored-by: default avatarDónal Murray <[email protected]>
Co-authored-by: default avatargupnik <[email protected]>
      bb8ddc46
  7. Jan 24, 2024
  9. Jan 22, 2024
  11. Jan 18, 2024
  13. Dec 21, 2023
  15. Nov 28, 2023
  17. Nov 10, 2023
  19. Oct 24, 2023
    • Kian Paimani's avatar
      Ensure correct variant count in `Runtime[Hold/Freeze]Reason` (#1900) · 35eb133b
      Kian Paimani authored 
      

closes https://github.com/paritytech/polkadot-sdk/issues/1882

## Breaking Changes

This PR introduces a new item to `pallet_balances::Config`:

```diff
trait Config {
++    type RuntimeFreezeReasons;
}
```

This value is only used to check it against `type MaxFreeze`. A similar
check has been added for `MaxHolds` against `RuntimeHoldReasons`, which
is already given to `pallet_balances`.

In all contexts, you should pass the real `RuntimeFreezeReasons`
generated by `construct_runtime` to `type RuntimeFreezeReasons`. Passing
`()` would also work, but it would imply that the runtime uses no
freezes at all.

---------

Signed-off-by: default avatarOliver Tale-Yazdi <[email protected]>
Co-authored-by: default avatarOliver Tale-Yazdi <[email protected]>
      35eb133b
  21. Oct 15, 2023
    • Daan van der Plas's avatar
      fix: GoAhead signal only set when runtime upgrade is enacted from parachain side (#1176) · 91c4360c
      Daan van der Plas authored 
      The runtime code of a parachain can be replaced on the relay-chain via:

[cumulus]:
[enact_authorized_upgrade](https://github.com/paritytech/polkadot-sdk/blob/1a38d6d6/cumulus/pallets/parachain-system/src/lib.rs#L661);
this is used for a runtime upgrade when a parachain is not bricked.

[polkadot] (these are used when a parachain is bricked):
-
[force_set_current_code](https://github.com/paritytech/polkadot-sdk/blob/1a38d6d6/polkadot/runtime/parachains/src/paras/mod.rs#L823):
immediately changes the runtime code of a given para without a pvf check
(root).
-
[force_schedule_code_upgrade](https://github.com/paritytech/polkadot-sdk/blob/1a38d6d6/polkadot/runtime/parachains/src/paras/mod.rs#L864):
schedules a change to the runtime code of a given para including a pvf
check of the new code (root).
-
[schedule_code_upgrade](https://github.com/paritytech/polkadot-sdk/blob/1a38d6d6/polkadot/runtime/common/src/paras_registrar.rs#L395):
schedules a change to the runtime code of a given para including a pvf
check of the new code. Besides root, the parachain or parachain manager
can call this extrinsic given that the parachain is unlocked.

Polkadot signals a parachain to be ready for a runtime upgrade through
the
[GoAhead](https://github.com/paritytech/polkadot-sdk/blob/e4949344

/polkadot/primitives/src/v5/mod.rs#L1229)
signal.

When in cumulus `enact_authorized_upgrade` is executed, the same
underlying helper function of `force_schedule_code_upgrade` &
`schedule_code_upgrade`:
[schedule_code_upgrade](https://github.com/paritytech/polkadot/blob/09b61286da11921a3dda0a8e4015ceb9ef9cffca/runtime/parachains/src/paras/mod.rs#L1778),
is called on the relay-chain, which sets the `GoAhead` signal (if the
pvf is accepted).

If Cumulus receives the `GoAhead` signal from polkadot without having
the `PendingValidationCode` ready, it will panic
([ref](https://github.com/paritytech/polkadot/pull/7412)). For
`enact_authorized_upgrade` we know for sure the `PendingValidationCode`
is set. On the contrary, for `force_schedule_code_upgrade` &
`schedule_code_upgrade` this is not the case.

This PR includes a flag such that the `GoAhead` signal will only be set
when a runtime upgrade is enacted by the parachain
(`enact_authorized_upgrade`).

additional info: https://github.com/paritytech/polkadot/pull/7412

Closes #641

---------

Co-authored-by: default avatarBastian Köcher <[email protected]>
Co-authored-by: default avatarBastian Köcher <[email protected]>
      91c4360c
  23. Sep 13, 2023
    • Liam Aharon's avatar
      Stabilize `VersionedMigration` (#1503) · 72de70c7
      Liam Aharon authored 
      `VersionedMigration` has become somewhat widely used for handling
version bumps in migrations the last few months.

It is currently behind the `experimental` feature flag, requiring every
pallet that writes a new migration with version bumps to set up the
`experimental` flag in their own Cargo.tomls, and also for every runtime
using these pallets to explicitly enable the `experimental` flag for
each pallet.

This is becoming quite verbose, and I can only see the number of pallets
requiring the experimental flag increasing for no other reason than
using what has become a commonly used feature.

Additionally, I'm writing migration docs and would like to avoid
stepping through how to use the `experimental` feature to get
`VersionedMigration` working.

Since the feature has been used in production for some time now without
any reported issues, is becoming commonly used and ready to advertise in
docs, I feel this is a good time to make it non-experimental.
      72de70c7
  25. Sep 06, 2023