add option to collect light client data (#3474)

Light clients require full nodes to serve additional data so that they
can stay in sync with the network. This patch adds a new launch option
`--import-light-client-data` to configure what data to make available.
For now, data is only kept in memory; it is not persisted at this time.
Note that data is only locally collected, a separate patch is needed to
actually make it availble over the network. `--serve-light-client-data`
will be used for serving data, but is not functional yet outside tests.

AllTests-mainnet.md

@@ -252,6 +252,13 @@ OK: 3/3 Fail: 0/3 Skip: 0/3
 + [SCRYPT] Network Keystore encryption                                                       OK
 ```
 OK: 9/9 Fail: 0/9 Skip: 0/9
+## Light client [Preset: mainnet]
+```diff
++ Init from checkpoint                                                                       OK
++ Light client sync                                                                          OK
++ Pre-Altair                                                                                 OK
+```
+OK: 3/3 Fail: 0/3 Skip: 0/3
 ## ListKeys requests [Preset: mainnet]
 ```diff
 + Correct token provided [Preset: mainnet]                                                   OK
@@ -513,4 +520,4 @@ OK: 1/1 Fail: 0/1 Skip: 0/1
 OK: 1/1 Fail: 0/1 Skip: 0/1
 
 ---TOTAL---
-OK: 282/287 Fail: 0/287 Skip: 5/287
+OK: 285/290 Fail: 0/290 Skip: 5/290

beacon_chain/conf.nim

@@ -28,6 +28,9 @@ import
   ./validators/slashing_protection_common,
   ./filepath
 
+from consensus_object_pools/block_pools_types_light_client
+  import ImportLightClientData
+
 export
   uri, nat, enr,
   defaultEth2TcpPort, enabledLogLevel, ValidIpAddress,
@@ -419,6 +422,19 @@ type
         desc: "A file specifying the authorization token required for accessing the keymanager API"
         name: "keymanager-token-file" }: Option[InputFile]
 
+      serveLightClientData* {.
+        hidden
+        desc: "BETA: Serve data for enabling light clients to stay in sync with the network"
+        defaultValue: false
+        name: "serve-light-client-data"}: bool
+
+      importLightClientData* {.
+        hidden
+        desc: "BETA: Which classes of light client data to import. " &
+              "Must be one of: none, only-new, full (slow startup), on-demand (may miss validator duties)"
+        defaultValue: ImportLightClientData.None
+        name: "import-light-client-data"}: ImportLightClientData
+
       inProcessValidators* {.
         desc: "Disable the push model (the beacon node tells a signing process with the private keys of the validators what to sign and when) and load the validators in the beacon node itself"
         defaultValue: true # the use of the nimbus_signing_process binary by default will be delayed until async I/O over stdin/stdout is developed for the child process.

beacon_chain/consensus_object_pools/block_clearance.nim

@@ -11,7 +11,7 @@ import
   chronicles,
   stew/[assign2, results],
   ../spec/[forks, signatures, signatures_batch, state_transition],
-  "."/[block_dag, blockchain_dag]
+  "."/[block_dag, blockchain_dag, blockchain_dag_light_client]
 
 export results, signatures_batch, block_dag, blockchain_dag
 
@@ -84,6 +84,9 @@ proc addResolvedHeadBlock(
     putBlockDur = putBlockTick - startTick,
     epochRefDur = epochRefTick - putBlockTick
 
+  # Update light client data
+  dag.processNewBlockForLightClient(state, trustedBlock, parent)
+
   # Notify others of the new block before processing the quarantine, such that
   # notifications for parents happens before those of the children
   if onBlockAdded != nil:

beacon_chain/consensus_object_pools/block_pools_types.nim

@@ -17,11 +17,11 @@ import
   ../spec/datatypes/[phase0, altair, bellatrix],
   ".."/beacon_chain_db,
   ../validators/validator_monitor,
-  ./block_dag
+  ./block_dag, block_pools_types_light_client
 
 export
   options, sets, tables, hashes, helpers, beacon_chain_db, block_dag,
-  validator_monitor
+  block_pools_types_light_client, validator_monitor
 
 # ChainDAG and types related to forming a DAG of blocks, keeping track of their
 # relationships and allowing various forms of lookups
@@ -178,6 +178,12 @@ type
 
     cfg*: RuntimeConfig
 
+    serveLightClientData*: bool
+      ## Whether to make local light client data available or not
+
+    importLightClientData*: ImportLightClientData
+      ## Which classes of light client data to import
+
     epochRefs*: array[32, EpochRef]
       ## Cached information about a particular epoch ending with the given
       ## block - we limit the number of held EpochRefs to put a cap on
@@ -189,6 +195,14 @@ type
       ## value with other components which don't have access to the
       ## full ChainDAG.
 
+    # -----------------------------------
+    # Data to enable light clients to stay in sync with the network
+
+    lightClientCache*: LightClientCache
+
+    # -----------------------------------
+    # Callbacks
+
     onBlockAdded*: OnBlockCallback
       ## On block added callback
     onHeadChanged*: OnHeadCallback
@@ -197,6 +211,8 @@ type
       ## On beacon chain reorganization
     onFinHappened*: OnFinalizedCallback
       ## On finalization callback
+    onOptimisticLightClientUpdate*: OnOptimisticLightClientUpdateCallback
+      ## On `OptimisticLightClientUpdate` updated callback
 
     headSyncCommittees*: SyncCommitteeCache
       ## A cache of the sync committees, as they appear in the head state -

beacon_chain/consensus_object_pools/block_pools_types_light_client.nim

@@ -0,0 +1,94 @@
+# beacon_chain
+# Copyright (c) 2022 Status Research & Development GmbH
+# Licensed and distributed under either of
+#   * MIT license (license terms in the root directory or at https://opensource.org/licenses/MIT).
+#   * Apache v2 license (license terms in the root directory or at https://www.apache.org/licenses/LICENSE-2.0).
+# at your option. This file may not be copied, modified, or distributed except according to those terms.
+
+{.push raises: [Defect].}
+
+# This implements the pre-release proposal of the libp2p based light client sync
+# protocol. See https://github.com/ethereum/consensus-specs/pull/2802
+
+import
+  # Status libraries
+  stew/bitops2,
+  # Beacon chain internals
+  ../spec/datatypes/altair,
+  ./block_dag
+
+type
+  OnOptimisticLightClientUpdateCallback* =
+    proc(data: OptimisticLightClientUpdate) {.gcsafe, raises: [Defect].}
+
+  ImportLightClientData* {.pure.} = enum
+    ## Controls which classes of light client data are imported.
+    None = "none"
+      ## Import no light client data.
+    OnlyNew = "only-new"
+      ## Import only new light client data.
+    Full = "full"
+      ## Import light client data for entire weak subjectivity period.
+    OnDemand = "on-demand"
+      ## No precompute of historic data. Is slow and may miss validator duties.
+
+  CachedLightClientData* = object
+    ## Cached data from historical non-finalized states to improve speed when
+    ## creating future `LightClientUpdate` and `LightClientBootstrap` instances.
+    current_sync_committee_branch*:
+      array[log2trunc(altair.CURRENT_SYNC_COMMITTEE_INDEX), Eth2Digest]
+
+    next_sync_committee_branch*:
+      array[log2trunc(altair.NEXT_SYNC_COMMITTEE_INDEX), Eth2Digest]
+
+    finalized_bid*: BlockId
+    finality_branch*:
+      array[log2trunc(altair.FINALIZED_ROOT_INDEX), Eth2Digest]
+
+  CachedLightClientBootstrap* = object
+    ## Cached data from historical finalized epoch boundary blocks to improve
+    ## speed when creating future `LightClientBootstrap` instances.
+    current_sync_committee_branch*:
+      array[log2trunc(altair.CURRENT_SYNC_COMMITTEE_INDEX), Eth2Digest]
+
+  LightClientCache* = object
+    data*: Table[BlockId, CachedLightClientData]
+      ## Cached data for creating future `LightClientUpdate` instances.
+      ## Key is the block ID of which the post state was used to get the data.
+      ## Data is stored for the most recent 4 finalized checkpoints, as well as
+      ## for all non-finalized blocks.
+
+    bootstrap*: Table[Slot, CachedLightClientBootstrap]
+      ## Cached data for creating future `LightClientBootstrap` instances.
+      ## Key is the block slot of which the post state was used to get the data.
+      ## Data is stored for finalized epoch boundary blocks.
+
+    latestCheckpoints*: array[4, Checkpoint]
+      ## Keeps track of the latest four `finalized_checkpoint` references
+      ## leading to `finalizedHead`. Used to prune `data`.
+      ## Non-finalized states may only refer to these checkpoints.
+
+    lastCheckpointIndex*: int
+      ## Last index that was modified in `latestCheckpoints`.
+
+    bestUpdates*: Table[SyncCommitteePeriod, altair.LightClientUpdate]
+      ## Stores the `LightClientUpdate` with the most `sync_committee_bits` per
+      ## `SyncCommitteePeriod`. Updates with a finality proof have precedence.
+
+    pendingBestUpdates*:
+      Table[(SyncCommitteePeriod, Eth2Digest), altair.LightClientUpdate]
+      ## Same as `bestUpdates`, but for `SyncCommitteePeriod` with
+      ## `next_sync_committee` that are not finalized. Key is `(period,
+      ## hash_tree_root(current_sync_committee | next_sync_committee)`.
+
+    latestUpdate*: altair.LightClientUpdate
+      ## Tracks the `LightClientUpdate` for the latest slot. This may be older
+      ## than head for empty slots or if not signed by sync committee.
+
+    optimisticUpdate*: OptimisticLightClientUpdate
+      ## Tracks the `OptimisticLightClientUpdate` for the latest slot. This may
+      ## be older than head for empty slots or if not signed by sync committee.
+
+    importTailSlot*: Slot
+      ## The earliest slot for which light client data is collected.
+      ## Only relevant for `ImportLightClientData.OnlyNew`.

beacon_chain/consensus_object_pools/blockchain_dag.nim

@@ -477,11 +477,23 @@ proc updateBeaconMetrics(state: StateData, cache: var StateCache) =
     beacon_active_validators.set(active_validators)
     beacon_current_active_validators.set(active_validators)
 
+import blockchain_dag_light_client
+
+export
+  blockchain_dag_light_client.getBestLightClientUpdateForPeriod,
+  blockchain_dag_light_client.getLatestLightClientUpdate,
+  blockchain_dag_light_client.getOptimisticLightClientUpdate,
+  blockchain_dag_light_client.getLightClientBootstrap
+
 proc init*(T: type ChainDAGRef, cfg: RuntimeConfig, db: BeaconChainDB,
            validatorMonitor: ref ValidatorMonitor, updateFlags: UpdateFlags,
            onBlockCb: OnBlockCallback = nil, onHeadCb: OnHeadCallback = nil,
            onReorgCb: OnReorgCallback = nil,
-           onFinCb: OnFinalizedCallback = nil): ChainDAGRef =
+           onFinCb: OnFinalizedCallback = nil,
+           onOptimisticLCUpdateCb: OnOptimisticLightClientUpdateCallback = nil,
+           serveLightClientData = false,
+           importLightClientData = ImportLightClientData.None
+          ): ChainDAGRef =
   # TODO we require that the db contains both a head and a tail block -
   #      asserting here doesn't seem like the right way to go about it however..
 
@@ -653,11 +665,14 @@ proc init*(T: type ChainDAGRef, cfg: RuntimeConfig, db: BeaconChainDB,
     # allow skipping some validation.
     updateFlags: {verifyFinalization} * updateFlags,
     cfg: cfg,
+    serveLightClientData: serveLightClientData,
+    importLightClientData: importLightClientData,
 
     onBlockAdded: onBlockCb,
     onHeadChanged: onHeadCb,
     onReorgHappened: onReorgCb,
-    onFinHappened: onFinCb
+    onFinHappened: onFinCb,
+    onOptimisticLightClientUpdate: onOptimisticLCUpdateCb
   )
 
   block: # Initialize dag states
@@ -779,6 +794,8 @@ proc init*(T: type ChainDAGRef, cfg: RuntimeConfig, db: BeaconChainDB,
     stateDur = stateTick - summariesTick,
     indexDur = Moment.now() - stateTick
 
+  dag.initLightClientCache()
+
   dag
 
 template genesisValidatorsRoot*(dag: ChainDAGRef): Eth2Digest =
@@ -1286,6 +1303,9 @@ proc pruneBlocksDAG(dag: ChainDAGRef) =
 
     var cur = head.atSlot()
     while not cur.blck.isAncestorOf(dag.finalizedHead.blck):
+      # Update light client data
+      dag.deleteLightClientData(cur.blck.bid)
+
       dag.delState(cur) # TODO: should we move that disk I/O to `onSlotEnd`
 
       if cur.isProposed():
@@ -1485,6 +1505,9 @@ proc updateHead*(
   doAssert (not finalizedHead.blck.isNil),
     "Block graph should always lead to a finalized block"
 
+  # Update light client data
+  dag.processHeadChangeForLightClient()
+
   let (isAncestor, ancestorDepth) = lastHead.getDepth(newHead)
   if not(isAncestor):
     notice "Updated head block with chain reorg",
@@ -1571,6 +1594,9 @@ proc updateHead*(
     # therefore no longer be considered as part of the chain we're following
     dag.pruneBlocksDAG()
 
+    # Update light client data
+    dag.processFinalizationForLightClient()
+
     # Send notification about new finalization point via callback.
     if not(isNil(dag.onFinHappened)):
       let stateRoot =