Add the genesis hash to the database (#2928)

Close #476

Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>

bin/light-base/src/json_rpc_service.rs

@@ -346,7 +346,7 @@ impl ServicePrototype {
                     Default::default(),
                 ),
             }),
-            genesis_block: config.genesis_block_hash,
+            genesis_block_hash: config.genesis_block_hash,
             next_subscription_id: atomic::AtomicU64::new(0),
             subscriptions: Mutex::new(Subscriptions {
                 misc: HashMap::with_capacity_and_hasher(
@@ -466,7 +466,7 @@ struct Background<TPlat: Platform> {
     /// Hash of the genesis block.
     /// Keeping the genesis block is important, as the genesis block hash is included in
     /// transaction signatures, and must therefore be queried by upper-level UIs.
-    genesis_block: [u8; 32],
+    genesis_block_hash: [u8; 32],
 
     /// Identifier to use for the next subscription.
     ///

bin/light-base/src/json_rpc_service/chain_head.rs

@@ -1441,6 +1441,7 @@ impl<TPlat: Platform> Background<TPlat> {
         let response = super::super::encode_database(
             &self.network_service.0,
             &self.sync_service,
+            &self.genesis_block_hash,
             self.sync_service.block_number_bytes(),
             usize::try_from(max_size_bytes.unwrap_or(u64::max_value()))
                 .unwrap_or(usize::max_value()),

bin/light-base/src/json_rpc_service/getters.rs

@@ -61,7 +61,7 @@ impl<TPlat: Platform> Background<TPlat> {
             .respond(
                 state_machine_request_id,
                 methods::Response::chainHead_unstable_genesisHash(methods::HashHexString(
-                    self.genesis_block,
+                    self.genesis_block_hash,
                 ))
                 .to_json_response(request_id),
             )
@@ -93,7 +93,7 @@ impl<TPlat: Platform> Background<TPlat> {
             .respond(
                 state_machine_request_id,
                 methods::Response::chainSpec_unstable_genesisHash(methods::HashHexString(
-                    self.genesis_block,
+                    self.genesis_block_hash,
                 ))
                 .to_json_response(request_id),
             )

bin/light-base/src/json_rpc_service/state_chain.rs

@@ -211,7 +211,7 @@ impl<TPlat: Platform> Background<TPlat> {
         let response = {
             match height {
                 Some(0) => methods::Response::chain_getBlockHash(methods::HashHexString(
-                    self.genesis_block,
+                    self.genesis_block_hash,
                 ))
                 .to_json_response(request_id),
                 None => {

bin/light-base/src/lib.rs

@@ -369,8 +369,16 @@ impl<TPlat: platform::Platform, TChain> Client<TPlat, TChain> {
                 ),
             ) {
                 // Use the database if it contains a more recent block than the chain spec checkpoint.
-                (Ok(Ok(genesis_ci)), checkpoint, Ok((database, checkpoint_nodes)))
-                    if checkpoint
+                (
+                    Ok(Ok(genesis_ci)),
+                    checkpoint,
+                    Ok((genesis_hash, database, checkpoint_nodes)),
+                ) if genesis_hash
+                    == genesis_ci
+                        .as_ref()
+                        .finalized_block_header
+                        .hash(chain_spec.block_number_bytes().into())
+                    && checkpoint
                         .as_ref()
                         .and_then(|r| r.as_ref().ok())
                         .map_or(true, |cp| {
@@ -386,7 +394,7 @@ impl<TPlat: platform::Platform, TChain> Client<TPlat, TChain> {
                 (
                     Err(chain_spec::FromGenesisStorageError::UnknownStorageItems),
                     checkpoint,
-                    Ok((database, checkpoint_nodes)),
+                    Ok((database_genesis_hash, database, checkpoint_nodes)),
                 ) if checkpoint
                     .as_ref()
                     .and_then(|r| r.as_ref().ok())
@@ -403,7 +411,20 @@ impl<TPlat: platform::Platform, TChain> Client<TPlat, TChain> {
                         digest: header::DigestRef::empty().into(),
                     };
 
-                    (database, genesis_header, checkpoint_nodes)
+                    if database_genesis_hash
+                        == genesis_header.hash(chain_spec.block_number_bytes().into())
+                    {
+                        (database, genesis_header, checkpoint_nodes)
+                    } else if let Some(Ok(checkpoint)) = checkpoint {
+                        // Database is incorrect.
+                        (checkpoint, genesis_header, checkpoint_nodes)
+                    } else {
+                        // TODO: we can in theory support chain specs that have neither a checkpoint nor the genesis storage, but it's complicated
+                        return Err(
+                            "Either a checkpoint or the genesis storage must be provided"
+                                .to_string(),
+                        );
+                    }
                 }
 
                 (Err(chain_spec::FromGenesisStorageError::UnknownStorageItems), None, _) => {
@@ -1115,11 +1136,13 @@ async fn start_services<TPlat: platform::Platform>(
 async fn encode_database<TPlat: platform::Platform>(
     network_service: &network_service::NetworkService<TPlat>,
     sync_service: &sync_service::SyncService<TPlat>,
+    genesis_block_hash: &[u8; 32],
     block_number_bytes: usize,
     max_size: usize,
 ) -> String {
     // Craft the structure containing all the data that we would like to include.
     let mut database_draft = SerdeDatabase {
+        genesis_hash: hex::encode(genesis_block_hash),
         chain: match sync_service.serialize_chain_information().await {
             Some(ci) => {
                 let encoded = finalized_serialize::encode_chain(&ci, block_number_bytes);
@@ -1188,13 +1211,20 @@ fn decode_database(
     block_number_bytes: usize,
 ) -> Result<
     (
+        [u8; 32],
         chain::chain_information::ValidChainInformation,
         Vec<(PeerId, Vec<multiaddr::Multiaddr>)>,
     ),
     (),
 > {
     let decoded: SerdeDatabase = serde_json::from_str(encoded).map_err(|_| ())?;
 
+    let genesis_hash = if decoded.genesis_hash.len() == 64 {
+        <[u8; 32]>::try_from(hex::decode(&decoded.genesis_hash).map_err(|_| ())?).unwrap()
+    } else {
+        return Err(());
+    };
+
     let (chain, _) = finalized_serialize::decode_chain(
         &serde_json::to_string(&decoded.chain).unwrap(),
         block_number_bytes,
@@ -1216,11 +1246,14 @@ fn decode_database(
         })
         .collect::<Vec<_>>();
 
-    Ok((chain, nodes))
+    Ok((genesis_hash, chain, nodes))
 }
 
 #[derive(serde::Serialize, serde::Deserialize)]
 struct SerdeDatabase {
+    /// Hexadecimal-encoded hash of the genesis block header. Has no `0x` prefix.
+    #[serde(rename = "genesisHash")]
+    genesis_hash: String,
     chain: Box<serde_json::value::RawValue>,
     nodes: hashbrown::HashMap<String, Vec<String>, fnv::FnvBuildHasher>,
 }

bin/wasm-node/CHANGELOG.md

@@ -6,6 +6,7 @@
 
 - The `payment_queryInfo` JSON-RPC function now works with runtimes that have defined the type of `Balance` to be less than 16 bytes. ([#2914](https://github.com/paritytech/smoldot/pull/2914))
 - The parameter of `chainHead_unstable_finalizedDatabase` has been renamed from `max_size_bytes` to `maxSizeBytes`. ([#2923](https://github.com/paritytech/smoldot/pull/2923))
+- The database now contains the hash of the genesis block header. This hash is verified when the database is loaded, and the database is ignored if there is a mismatch. This prevents accidents where the wrong database is provided, which would lead to the chain not working and would be hard to debug. ([#2928](https://github.com/paritytech/smoldot/pull/2928))
 
 ## 0.7.3 - 2022-10-19