Fix missing checkpoint states` (#3225)

With the right sequence of events (for example a REST request or a
validation), it can happen that the first traversal across a state
checkpoint boundary is done without storing that state on disk - this
causes problens when replaying states, because now states may be missing
from the database.

Here, we simply avoid using the caches when advancing a state that will
go into the database, ensuring that the information lost during caching
always is permanently stored.

* fix recursion bug in `isProposed`

beacon_chain/consensus_object_pools/block_clearance.nim

@@ -222,6 +222,12 @@ proc addHeadBlock*(
   # blocks we add to the database are clean for the given state
   let startTick = Moment.now()
 
+  # The clearance state works as the canonical
+  # "let's make things permanent" point and saves things to the database -
+  # storing things is slow, so we don't want to do so before there's a
+  # reasonable chance that the information will become more permanently useful -
+  # by the time a new block reaches this point, the parent block will already
+  # have "established" itself in the network to some degree at least.
   var cache = StateCache()
   updateStateData(
     dag, dag.clearanceState, parent.atSlot(signedBlock.message.slot), true, cache)

beacon_chain/consensus_object_pools/block_dag.nim

@@ -201,7 +201,7 @@ func isProposed*(bid: BlockId, slot: Slot): bool =
 
 func isProposed*(blck: BlockRef, slot: Slot): bool =
   ## Return true if `blck` was proposed in the given slot
-  not isNil(blck) and blck.isProposed(slot)
+  not isNil(blck) and blck.bid.isProposed(slot)
 
 func isProposed*(bs: BlockSlot): bool =
   ## Return true if `bs` represents the proposed block (as opposed to an empty

beacon_chain/consensus_object_pools/blockchain_dag.nim

@@ -792,6 +792,7 @@ proc advanceSlots(
   # processing - the state must be positions at a slot before or equal to the
   # target
   doAssert getStateField(state.data, slot) <= slot
+
   while getStateField(state.data, slot) < slot:
     let preEpoch = getStateField(state.data, slot).epoch
     loadStateCache(dag, cache, state.blck, getStateField(state.data, slot).epoch)
@@ -874,7 +875,11 @@ proc updateStateData*(
   # it also avoids `hash_tree_root` for slot processing
   if exactMatch(state, cur):
     found = true
-  elif exactMatch(dag.headState, cur):
+  elif not save:
+    # When required to save states, we cannot rely on the caches because that
+    # would skip the extra processing that save does - not all information that
+    # goes into the database is cached
+    if exactMatch(dag.headState, cur):
       assign(state, dag.headState)
       found = true
     elif exactMatch(dag.clearanceState, cur):
@@ -893,10 +898,11 @@ proc updateStateData*(
   # sequentially to grab their votes.
   const RewindBlockThreshold = 64
   while not found and ancestors.len < RewindBlockThreshold:
-    if canAdvance(state, cur):
+    if canAdvance(state, cur): # Typical case / fast path when there's no reorg
       found = true
       break
 
+    if not save: # See above
       if canAdvance(dag.headState, cur):
         assign(state, dag.headState)
         found = true

tests/test_block_dag.nim

@@ -87,6 +87,9 @@ suite "BlockSlot and helpers":
       s24.parent == BlockSlot(blck: s2, slot: Slot(3))
       s24.parent.parent == s22
 
+      s22.isProposed()
+      not s24.isProposed()
+
 suite "BlockId and helpers":
   test "atSlot sanity":
     let

tests/test_blockchain_dag.nim

@@ -160,6 +160,30 @@ suite "Block pool processing" & preset():
       dag.getBlockRange(Slot(3), 2, blocks.toOpenArray(0, 1)) == 2
       blocks[2..<2].len == 0
 
+    # A fork forces the clearance state to a point where it cannot be advanced
+    let
+      nextEpoch = dag.head.slot.epoch + 1
+      nextEpochSlot = nextEpoch.compute_start_slot_at_epoch()
+      stateCheckpoint = dag.head.parent.atSlot(nextEpochSlot).stateCheckpoint
+
+    check:
+      not dag.getEpochRef(dag.head.parent, nextEpoch).isNil
+
+      # Getting an EpochRef should not result in states being stored
+      db.getStateRoot(stateCheckpoint.blck.root, stateCheckpoint.slot).isErr()
+      # this is required for the test to work - it's not a "public"
+      # post-condition of getEpochRef
+      getStateField(dag.epochRefState.data, slot) == nextEpochSlot
+    assign(state[], dag.epochRefState.data)
+
+    let
+      bnext = addTestBlock(state[], cache).phase0Data
+      bnextAdd = dag.addHeadBlock(verifier, bnext, nilPhase0Callback)
+
+    check:
+      # Getting an EpochRef should not result in states being stored
+      db.getStateRoot(stateCheckpoint.blck.root, stateCheckpoint.slot).isOk()
+
   test "Adding the same block twice returns a Duplicate error" & preset():
     let
       b10 = dag.addHeadBlock(verifier, b1, nilPhase0Callback)