XCM: SubscribeVersion doesn't work over bridge(s)
#2417
Assignees
Labels
I2-bug
The node fails to follow expected behavior.
T6-XCM
This PR/Issue is related to XCM.
T15-bridges
This PR/Issue is related to bridges.
Comments
acatangiu
added
T6-XCM
|
I am taking this also as a part of paritytech/parity-bridges-common#2500 |
This was referenced Nov 23, 2023
bkontur
added a commit
that referenced
this issue
Dec 12, 2023
## Summary This pull request proposes a solution for improved control of the versioned XCM flow over the bridge (across different consensus chains) and resolves the situation where the sending chain/consensus has already migrated to a higher XCM version than the receiving chain/consensus. ## Problem/Motivation The current flow over the bridge involves a transfer from AssetHubRococo (AHR) to BridgeHubRococo (BHR) to BridgeHubWestend (BHW) and finally to AssetHubWestend (AHW), beginning with a reserve-backed transfer on AHR. In this process: 1. AHR sends XCM `ExportMessage` through `XcmpQueue`, incorporating XCM version checks using the `WrapVersion` feature, influenced by `pallet_xcm::SupportedVersion` (managed by `pallet_xcm::force_xcm_version` or version discovery). 2. BHR handles the `ExportMessage` instruction, utilizing the latest XCM version. The `HaulBlobExporter` converts the inner XCM to [`VersionedXcm::from`](https://github.com/paritytech/polkadot-sdk/blob/63ac2471aa0210f0ac9903bdd7d8f9351f9a635f/polkadot/xcm/xcm-builder/src/universal_exports.rs#L465-L467), also using the latest XCM version. However, challenges arise: - Incompatibility when BHW uses a different version than BHR. For instance, if BHR migrates to **XCMv4** while BHW remains on **XCMv3**, BHR's `VersionedXcm::from` uses `VersionedXcm::V4` variant, causing encoding issues for BHW. ``` /// Just a simulation of possible error, which could happen on BHW /// (this code is based on actual master without XCMv4) let encoded = hex_literal::hex!("0400"); println!("{:?}", VersionedXcm::<()>::decode(&mut &encoded[..])); Err(Error { cause: None, desc: "Could not decode `VersionedXcm`, variant doesn't exist" }) ``` - Similar compatibility issues exist between AHR and AHW. ## Solution This pull request introduces the following solutions: 1. **New trait `CheckVersion`** - added to the `xcm` module and exposing `pallet_xcm::SupportedVersion`. This enhancement allows checking the actual XCM version for desired destinations outside of the `pallet_xcm` module. 2. **Version Check in `HaulBlobExporter`** uses `CheckVersion` to check known/configured destination versions, ensuring compatibility. For example, in the scenario mentioned, BHR can store the version `3` for BHW. If BHR is on XCMv4, it will attempt to downgrade the message to version `3` instead of using the latest version `4`. 3. **Version Check in `pallet-xcm-bridge-hub-router`** - this check ensures compatibility with the real destination's XCM version, preventing the unnecessary sending of messages to the local bridge hub if versions are incompatible. These additions aim to improve the control and compatibility of XCM flows over the bridge and addressing issues related to version mismatches. ## Possible alternative solution _(More investigation is needed, and at the very least, it should extend to XCMv4/5. If this proves to be a viable option, I can open an RFC for XCM.)._ Add the `XcmVersion` attribute to the `ExportMessage` so that the sending chain can determine, based on what is stored in `pallet_xcm::SupportedVersion`, the version the destination is using. This way, we may not need to handle the version in `HaulBlobExporter`. ``` ExportMessage { network: NetworkId, destination: InteriorMultiLocation, xcm: Xcm<()> destination_xcm_version: Version, // <- new attritbute }, ``` ``` pub trait ExportXcm { fn validate( network: NetworkId, channel: u32, universal_source: &mut Option<InteriorMultiLocation>, destination: &mut Option<InteriorMultiLocation>, message: &mut Option<Xcm<()>>, destination_xcm_version: Version, , // <- new attritbute ) -> SendResult<Self::Ticket>; ``` ## Future Directions This PR does not fix version discovery over bridge, further investigation will be conducted here: #2417. ## TODO - [x] `pallet_xcm` mock for tests uses hard-coded XCM version `2` - change to 3 or lastest? - [x] fix `pallet-xcm-bridge-hub-router` - [x] fix HaulBlobExporter with version determination [here](https://github.com/paritytech/polkadot-sdk/blob/2183669d05f9b510f979a0cc3c7847707bacba2e/polkadot/xcm/xcm-builder/src/universal_exports.rs#L465) - [x] add unit-tests to the runtimes - [x] run benchmarks for `ExportMessage` - [x] extend local run scripts about `force_xcm_version(dest, version)` - [ ] when merged, prepare governance calls for Rococo/Westend - [ ] add PRDoc Part of: paritytech/parity-bridges-common#2719 --------- Co-authored-by: command-bot <>
franciscoaguirre
pushed a commit
to paritytech/xcm
that referenced
this issue
Jan 25, 2024
## Summary This pull request proposes a solution for improved control of the versioned XCM flow over the bridge (across different consensus chains) and resolves the situation where the sending chain/consensus has already migrated to a higher XCM version than the receiving chain/consensus. ## Problem/Motivation The current flow over the bridge involves a transfer from AssetHubRococo (AHR) to BridgeHubRococo (BHR) to BridgeHubWestend (BHW) and finally to AssetHubWestend (AHW), beginning with a reserve-backed transfer on AHR. In this process: 1. AHR sends XCM `ExportMessage` through `XcmpQueue`, incorporating XCM version checks using the `WrapVersion` feature, influenced by `pallet_xcm::SupportedVersion` (managed by `pallet_xcm::force_xcm_version` or version discovery). 2. BHR handles the `ExportMessage` instruction, utilizing the latest XCM version. The `HaulBlobExporter` converts the inner XCM to [`VersionedXcm::from`](https://github.com/paritytech/polkadot-sdk/blob/63ac2471aa0210f0ac9903bdd7d8f9351f9a635f/polkadot/xcm/xcm-builder/src/universal_exports.rs#L465-L467), also using the latest XCM version. However, challenges arise: - Incompatibility when BHW uses a different version than BHR. For instance, if BHR migrates to **XCMv4** while BHW remains on **XCMv3**, BHR's `VersionedXcm::from` uses `VersionedXcm::V4` variant, causing encoding issues for BHW. ``` /// Just a simulation of possible error, which could happen on BHW /// (this code is based on actual master without XCMv4) let encoded = hex_literal::hex!("0400"); println!("{:?}", VersionedXcm::<()>::decode(&mut &encoded[..])); Err(Error { cause: None, desc: "Could not decode `VersionedXcm`, variant doesn't exist" }) ``` - Similar compatibility issues exist between AHR and AHW. ## Solution This pull request introduces the following solutions: 1. **New trait `CheckVersion`** - added to the `xcm` module and exposing `pallet_xcm::SupportedVersion`. This enhancement allows checking the actual XCM version for desired destinations outside of the `pallet_xcm` module. 2. **Version Check in `HaulBlobExporter`** uses `CheckVersion` to check known/configured destination versions, ensuring compatibility. For example, in the scenario mentioned, BHR can store the version `3` for BHW. If BHR is on XCMv4, it will attempt to downgrade the message to version `3` instead of using the latest version `4`. 3. **Version Check in `pallet-xcm-bridge-hub-router`** - this check ensures compatibility with the real destination's XCM version, preventing the unnecessary sending of messages to the local bridge hub if versions are incompatible. These additions aim to improve the control and compatibility of XCM flows over the bridge and addressing issues related to version mismatches. ## Possible alternative solution _(More investigation is needed, and at the very least, it should extend to XCMv4/5. If this proves to be a viable option, I can open an RFC for XCM.)._ Add the `XcmVersion` attribute to the `ExportMessage` so that the sending chain can determine, based on what is stored in `pallet_xcm::SupportedVersion`, the version the destination is using. This way, we may not need to handle the version in `HaulBlobExporter`. ``` ExportMessage { network: NetworkId, destination: InteriorMultiLocation, xcm: Xcm<()> destination_xcm_version: Version, // <- new attritbute }, ``` ``` pub trait ExportXcm { fn validate( network: NetworkId, channel: u32, universal_source: &mut Option<InteriorMultiLocation>, destination: &mut Option<InteriorMultiLocation>, message: &mut Option<Xcm<()>>, destination_xcm_version: Version, , // <- new attritbute ) -> SendResult<Self::Ticket>; ``` ## Future Directions This PR does not fix version discovery over bridge, further investigation will be conducted here: paritytech/polkadot-sdk#2417. ## TODO - [x] `pallet_xcm` mock for tests uses hard-coded XCM version `2` - change to 3 or lastest? - [x] fix `pallet-xcm-bridge-hub-router` - [x] fix HaulBlobExporter with version determination [here](https://github.com/paritytech/polkadot-sdk/blob/2183669d05f9b510f979a0cc3c7847707bacba2e/polkadot/xcm/xcm-builder/src/universal_exports.rs#L465) - [x] add unit-tests to the runtimes - [x] run benchmarks for `ExportMessage` - [x] extend local run scripts about `force_xcm_version(dest, version)` - [ ] when merged, prepare governance calls for Rococo/Westend - [ ] add PRDoc Part of: paritytech/parity-bridges-common#2719 --------- Co-authored-by: command-bot <>
10 tasks
|
@bkontur is this done? |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
originally posted by @jakoblell
I've just stumbled over a potential bug in the XCM executor which may become relevant for the bridge deployment: The
SubscribeVersioninstruction ensures thatoriginal_origin == originand this check will fail if the origin has been changed byUniversalOrigin/DescendOrigin(which will be the case for messages coming in via the bridge). So as of now it won't be possible to use this version subscription mechanism for bridges.Location of the check:
https://github.com/paritytech/polkadot-sdk/blob/master/polkadot/xcm/xcm-executor/src/lib.rs#L741
Please note that just removing this check would open up the chain for DoS, so maybe some form of hardcoded (or configurable by root) allowlist is required for that.
TODO
VersionDiscoveryQueue(e.g. trim location to relay or parachain level?)pallet-xcm-bridge-hub-routersee comments bellow https://github.com/paritytech/polkadot-sdk/pull/2481/files#r1414102227The text was updated successfully, but these errors were encountered: