add option to produce light client data

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
`--serve-light-client-data` to enable collection ot light client data.
Note that data is only produced locally, a separate patch is needed to
actually make the data availble over the network.

Author: etan-status

AllTests-mainnet.md

@@ -223,6 +223,12 @@ 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
++ Light client sync                                                                          OK
++ Pre-Altair                                                                                 OK
+```
+OK: 2/2 Fail: 0/2 Skip: 0/2
 ## ListKeys requests [Preset: mainnet]
 ```diff
 + Correct token provided [Preset: mainnet]                                                   OK
@@ -450,4 +456,4 @@ OK: 1/1 Fail: 0/1 Skip: 0/1
 OK: 1/1 Fail: 0/1 Skip: 0/1
 
 ---TOTAL---
-OK: 248/252 Fail: 0/252 Skip: 4/252
+OK: 250/254 Fail: 0/254 Skip: 4/254

beacon_chain/conf.nim

@@ -1,5 +1,5 @@
 # beacon_chain
-# Copyright (c) 2018-2021 Status Research & Development GmbH
+# Copyright (c) 2018-2022 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).
@@ -360,6 +360,11 @@ type
         desc: "A file specifying the authorizition token required for accessing the keymanager API"
         name: "keymanager-token-file" }: Option[InputFile]
 
+      serveLightClientData* {.
+        desc: "Serve data to allow light clients to sync"
+        defaultValue: false
+        name: "serve-light-client-data"}: bool
+
       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

@@ -1,5 +1,5 @@
 # beacon_chain
-# Copyright (c) 2018-2021 Status Research & Development GmbH
+# Copyright (c) 2018-2022 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).
@@ -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
 
@@ -65,10 +65,12 @@ proc addResolvedHeadBlock(
   # been applied but the `blck` field was still set to the parent
   state.blck = blockRef
 
+  # Enable light clients to stay in sync with the network.
+  dag.processNewBlockForLightClient(state, trustedBlock, parent)
+
   # Regardless of the chain we're on, the deposits come in the same order so
   # as soon as we import a block, we'll also update the shared public key
   # cache
-
   dag.updateValidatorKeys(getStateField(state.data, validators).asSeq())
 
   # Getting epochRef with the state will potentially create a new EpochRef

beacon_chain/consensus_object_pools/block_dag.nim

@@ -57,6 +57,12 @@ type
       ## Slot time for this BlockSlot which may differ from blck.slot when time
       ## has advanced without blocks
 
+func hash*(bid: BlockId): Hash =
+  var h: Hash = 0
+  h = h !& hash(bid.root)
+  h = h !& hash(bid.slot)
+  !$h
+
 template root*(blck: BlockRef): Eth2Digest = blck.bid.root
 template slot*(blck: BlockRef): Slot = blck.bid.slot
 

beacon_chain/consensus_object_pools/block_pools_types.nim

@@ -11,7 +11,7 @@ import
   # Standard library
   std/[options, sets, tables, hashes],
   # Status libraries
-  chronicles,
+  stew/[bitops2, endians2], chronicles,
   # Internals
   ../spec/[signatures_batch, forks, helpers],
   ../spec/datatypes/[phase0, altair, bellatrix],
@@ -52,6 +52,8 @@ type
     proc(data: ReorgInfoObject) {.gcsafe, raises: [Defect].}
   OnFinalizedCallback* =
     proc(data: FinalizationInfoObject) {.gcsafe, raises: [Defect].}
+  OnOptimisticLightClientUpdateCallback* =
+    proc(data: OptimisticLightClientUpdate) {.gcsafe, raises: [Defect].}
 
   FetchRecord* = object
     root*: Eth2Digest
@@ -63,6 +65,24 @@ type
     # unnecessary overhead.
     data*: BlockRef
 
+  CachedSyncCommittee* = object
+    ## Cached sync committee. Note that the next sync committee is not finalized
+    ## until the first block of the sync committee period is finalized. Forks at
+    ## the beginning of a new sync committee period may refer to different next
+    ## sync committees.
+    sync_committee*: SyncCommittee
+    ref_count*: uint64
+
+  CachedLightClientData* = object
+    ## Cached data from historical states to improve speed when creating
+    ## future `LightClientUpdate` instances.
+    next_sync_committee_root*: Eth2Digest
+    next_sync_committee_branch*:
+      array[log2trunc(NEXT_SYNC_COMMITTEE_INDEX), Eth2Digest]
+
+    finalized_bid*: BlockId
+    finality_branch*: array[log2trunc(FINALIZED_ROOT_INDEX), Eth2Digest]
+
   ChainDAGRef* = ref object
     ## Pool of blocks responsible for keeping a DAG of resolved blocks.
     ##
@@ -151,6 +171,9 @@ type
 
     cfg*: RuntimeConfig
 
+    createLightClientData*: bool
+      ## Whether or not `LightClientUpdate` should be produced.
+
     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
@@ -162,6 +185,41 @@ type
       ## value with other components which don't have access to the
       ## full ChainDAG.
 
+    # -----------------------------------
+    # Cached data to enable light clients to stay in sync with the network
+
+    cachedSyncCommittees*: Table[Eth2Digest, CachedSyncCommittee]
+      ## Cached sync committees for creating future `LightClientUpdate`
+      ## instances. Key is `hash_tree_root(sync_committee)`.
+      ## Count is reference count, from `cachedLightClientData`.
+
+    cachedLightClientData*: 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.
+
+    lightClientCheckpoints*: array[4, Checkpoint]
+      ## Keeps track of the latest four `finalized_checkpoint` references
+      ## leading to `finalizedHead`. Used to prune `cachedLightClientData`.
+      ## Non-finalized states may only refer to these checkpoints.
+
+    lastLightClientCheckpointIndex*: int
+      ## Last index that was modified in `lightClientCheckpoints`.
+
+    bestLightClientUpdates*: Table[SyncCommitteePeriod, LightClientUpdate]
+      ## Stores the `LightClientUpdate` with the most `sync_committee_bits` per
+      ## `SyncCommitteePeriod`. Updates with a finality proof have precedence.
+
+    latestLightClientUpdate*: LightClientUpdate
+      ## Tracks the `LightClientUpdate` for the latest slot. This may be older
+      ## than head for empty slots or if not signed by sync committee.
+
+    optimisticLightClientUpdate*: OptimisticLightClientUpdate
+      ## Tracks the `OptimisticLightClientUpdate` for the latest slot. This may
+      ## be older than head for empty slots or if not signed by sync committee.
+
+    # -----------------------------------
+    # Callbacks
+
     onBlockAdded*: OnBlockCallback
       ## On block added callback
     onHeadChanged*: OnHeadCallback
@@ -170,6 +228,8 @@ 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/blockchain_dag.nim

@@ -8,12 +8,12 @@
 {.push raises: [Defect].}
 
 import
-  std/[options, sequtils, tables, sets],
-  stew/[assign2, byteutils, results],
+  std/[algorithm, options, sequtils, tables, sets],
+  stew/[assign2, bitops2, byteutils, objects, results],
   metrics, snappy, chronicles,
   ../spec/[beaconstate, eth2_merkleization, eth2_ssz_serialization, helpers,
     state_transition, validator],
-  ../spec/datatypes/[phase0, altair],
+  ../spec/datatypes/[phase0, altair, bellatrix],
   ".."/beacon_chain_db,
   "."/[block_pools_types, block_quarantine]
 
@@ -53,7 +53,8 @@ const
   EPOCHS_PER_STATE_SNAPSHOT = 32
 
 proc putBlock*(
-    dag: ChainDAGRef, signedBlock: ForkyTrustedSignedBeaconBlock) =
+    dag: ChainDAGRef,
+    signedBlock: ForkyTrustedSignedBeaconBlock) =
   dag.db.putBlock(signedBlock)
 
 proc updateStateData*(
@@ -401,11 +402,24 @@ proc getForkedBlock*(
   dag.getForkedBlock(blck.bid).expect(
     "BlockRef block should always load, database corrupt?")
 
+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,
+           createLightClientData = false): ChainDAGRef =
+  if onOptimisticLCUpdateCb != nil:
+    doAssert createLightClientData
+
   # 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..
 
@@ -543,6 +557,7 @@ proc init*(T: type ChainDAGRef, cfg: RuntimeConfig, db: BeaconChainDB,
     # allow skipping some validation.
     updateFlags: {verifyFinalization} * updateFlags,
     cfg: cfg,
+    createLightClientData: createLightClientData,
 
     forkDigests: newClone ForkDigests.init(
       cfg,
@@ -551,7 +566,8 @@ proc init*(T: type ChainDAGRef, cfg: RuntimeConfig, db: BeaconChainDB,
     onBlockAdded: onBlockCb,
     onHeadChanged: onHeadCb,
     onReorgHappened: onReorgCb,
-    onFinHappened: onFinCb
+    onFinHappened: onFinCb,
+    onOptimisticLightClientUpdate: onOptimisticLCUpdateCb
   )
 
   let forkVersions =
@@ -618,6 +634,84 @@ proc init*(T: type ChainDAGRef, cfg: RuntimeConfig, db: BeaconChainDB,
     forkBlocks = dag.forkBlocks.len(),
     backfill = (dag.backfill.slot, shortLog(dag.backfill.parent_root))
 
+  # Initialize cached light client data (finalized head + non-finalized blocks).
+  if dag.createLightClientData:
+    let altairStartSlot = dag.cfg.ALTAIR_FORK_EPOCH.start_slot
+    if headRef.slot >= altairStartSlot:
+      let
+        finalizedSlot = dag.finalizedHead.blck.slot
+        finalizedPeriod = finalizedSlot.sync_committee_period
+      dag.initializeBestLightClientUpdateForPeriod(
+        finalizedPeriod, prune = false)
+
+      debug "Initializing cached light client data"
+      let lightClientStartTick = Moment.now()
+
+      # Build lists of block to process.
+      # As it is slow to load states in descending order,
+      # first build a todo list, then process them in ascending order.
+      let earliestSlot = max(finalizedSlot, altairStartSlot)
+      var
+        blocksBetween = newSeqOfCap[BlockRef](headRef.slot - earliestSlot + 1)
+        blockRef = headRef
+      while blockRef.slot > earliestSlot:
+        blocksBetween.add blockRef
+        blockRef = blockRef.parent
+      blocksBetween.add blockRef
+
+      # Process blocks.
+      let lowSlot = max(altairStartSlot, dag.tail.slot)
+      var
+        oldCheckpoint: Checkpoint
+        checkpointIndex = 0
+      for i in countdown(blocksBetween.high, blocksBetween.low):
+        blockRef = blocksBetween[i]
+        dag.withUpdatedState(dag.headState, blockRef.atSlot(blockRef.slot)) do:
+          withStateAndBlck(stateData.data, dag.getForkedBlock(blck)):
+            when stateFork >= BeaconStateFork.Altair:
+              # Cache data for `LightClientUpdate` of descendant blocks.
+              dag.cacheLightClientData(state, blck, isNew = false)
+
+              # Cache data for the block's `finalized_checkpoint`.
+              # The `finalized_checkpoint` may refer to:
+              # 1. `finalizedHead.blck -> finalized_checkpoint`
+              #    This may happen when there were skipped slots.
+              # 2. `finalizedHead -> finalized_checkpoint`
+              # 3. One epoch boundary that got justified then finalized
+              #    between `finalizedHead -> finalized_checkpoint`
+              #    and `finalizedHead`
+              # 4. `finalizedHead`
+              let checkpoint = state.data.finalized_checkpoint
+              if checkpoint != oldCheckpoint:
+                oldCheckpoint = checkpoint
+                doAssert checkpointIndex < dag.lightClientCheckpoints.len
+                dag.lightClientCheckpoints[checkpointIndex] = checkpoint
+                dag.lastLightClientCheckpointIndex = checkpointIndex
+                inc checkpointIndex
+                if checkpoint.root != dag.finalizedHead.blck.root:
+                  let cpRef =
+                    dag.getBlockAtSlot(checkpoint.epoch.start_slot).blck
+                  if cpRef != nil and cpRef.slot >= lowSlot:
+                    assert cpRef.bid.root == checkpoint.root
+                    dag.withUpdatedState(tmpState[],
+                        cpRef.atSlot(cpRef.slot)) do:
+                      withStateAndBlck(
+                          stateData.data, dag.getForkedBlock(blck)):
+                        when stateFork >= BeaconStateFork.Altair:
+                          dag.cacheLightClientData(state, blck, isNew = false)
+                        else: raiseAssert "Unreachable"
+                    do: raiseAssert "Unreachable"
+
+              # Create `LightClientUpdate` for non-finalized blocks.
+              if blockRef.slot > earliestSlot:
+                dag.createLightClientUpdates(state, blck, blockRef.parent)
+            else: raiseAssert "Unreachable"
+        do: raiseAssert "Unreachable"
+
+      let lightClientEndTick = Moment.now()
+      debug "Initialized cached light client data",
+        initDur = lightClientEndTick - lightClientStartTick
+
   dag
 
 template genesisValidatorsRoot*(dag: ChainDAGRef): Eth2Digest =
@@ -1127,6 +1221,9 @@ proc pruneBlocksDAG(dag: ChainDAGRef) =
 
     var cur = head.atSlot()
     while not cur.blck.isAncestorOf(dag.finalizedHead.blck):
+      if dag.createLightClientData:
+        dag.deleteLightClientData(cur.blck.bid)
+
       dag.delState(cur) # TODO: should we move that disk I/O to `onSlotEnd`
 
       if cur.isProposed():
@@ -1447,6 +1544,8 @@ proc updateHead*(
     # in order to clear out blocks that are no longer viable and should
     # therefore no longer be considered as part of the chain we're following
     dag.pruneBlocksDAG()
+    if dag.createLightClientData:
+      dag.pruneLightClientData()
 
     # Send notification about new finalization point via callback.
     if not(isNil(dag.onFinHappened)):