Fix block backfill with genesis skip slots (#4820)

## Issue Addressed

Closes #4817.

## Proposed Changes

- Fill in the linear block roots array between 0 and the slot of the first block (e.g. slots 0 and 1 on Holesky).
- Backport the `--freezer`, `--skip` and `--limit` options for `lighthouse db inspect` from tree-states. This allows us to easily view the database corruption of 4817 using `lighthouse db inspect --network holesky --freezer --column bbr --output values --limit 2`.
- Backport the `iter_column_from` change and `MemoryStore` overhaul from tree-states. These are required to enable `lighthouse db inspect`.
- Rework `freezer_upper_limit` to allow state lookups for slots below the `state_lower_limit`. Currently state lookups will fail until state reconstruction completes entirely.

There is a new regression test for the main bug, but no test for the `freezer_upper_limit` fix because we don't currently support running state reconstruction partially (see #3026). This will be fixed once we merge `tree-states`! In lieu of an automated test, I've tested manually on a Holesky node while it was reconstructing.

## Additional Info

Users who backfilled Holesky to slot 0 (e.g. using `--reconstruct-historic-states`) need to either:

- Re-sync from genesis.
- Re-sync using checkpoint sync and the changes from this PR.

Due to the recency of the Holesky genesis, writing a custom pass to fix up broken databases (which would require its own thorough testing) was deemed unnecessary. This is the primary reason for this PR being marked `backwards-incompat`.

This will create few conflicts with Deneb, which I've already resolved on `tree-states-deneb` and will be happy to backport to Deneb once this PR is merged to unstable.

beacon_node/store/src/errors.rs

@@ -45,6 +45,8 @@ pub enum Error {
     BlockReplayError(BlockReplayError),
     AddPayloadLogicError,
     SlotClockUnavailableForMigration,
+    InvalidKey,
+    InvalidBytes,
     UnableToDowngrade,
     InconsistentFork(InconsistentFork),
 }

beacon_node/store/src/hot_cold_store.rs

@@ -1490,10 +1490,17 @@ impl<E: EthSpec, Hot: ItemStore<E>, Cold: ItemStore<E>> HotColdDB<E, Hot, Cold>
         let split_slot = self.get_split_slot();
         let anchor = self.get_anchor_info();
 
-        // There are no restore points stored if the state upper limit lies in the hot database.
-        // It hasn't been reached yet, and may never be.
-        if anchor.map_or(false, |a| a.state_upper_limit >= split_slot) {
+        // There are no restore points stored if the state upper limit lies in the hot database,
+        // and the lower limit is zero. It hasn't been reached yet, and may never be.
+        if anchor.as_ref().map_or(false, |a| {
+            a.state_upper_limit >= split_slot && a.state_lower_limit == 0
+        }) {
             None
+        } else if let Some(lower_limit) = anchor
+            .map(|a| a.state_lower_limit)
+            .filter(|limit| *limit > 0)
+        {
+            Some(lower_limit)
         } else {
             Some(
                 (split_slot - 1) / self.config.slots_per_restore_point

beacon_node/store/src/leveldb_store.rs

@@ -1,7 +1,6 @@
 use super::*;
 use crate::hot_cold_store::HotColdDBError;
 use crate::metrics;
-use db_key::Key;
 use leveldb::compaction::Compaction;
 use leveldb::database::batch::{Batch, Writebatch};
 use leveldb::database::kv::KV;
@@ -176,10 +175,8 @@ impl<E: EthSpec> KeyValueStore<E> for LevelDB<E> {
         Ok(())
     }
 
-    /// Iterate through all keys and values in a particular column.
-    fn iter_column(&self, column: DBColumn) -> ColumnIter {
-        let start_key =
-            BytesKey::from_vec(get_key_for_col(column.into(), Hash256::zero().as_bytes()));
+    fn iter_column_from<K: Key>(&self, column: DBColumn, from: &[u8]) -> ColumnIter<K> {
+        let start_key = BytesKey::from_vec(get_key_for_col(column.into(), from));
 
         let iter = self.db.iter(self.read_options());
         iter.seek(&start_key);
@@ -187,13 +184,12 @@ impl<E: EthSpec> KeyValueStore<E> for LevelDB<E> {
         Box::new(
             iter.take_while(move |(key, _)| key.matches_column(column))
                 .map(move |(bytes_key, value)| {
-                    let key =
-                        bytes_key
-                            .remove_column(column)
-                            .ok_or(HotColdDBError::IterationError {
-                                unexpected_key: bytes_key,
-                            })?;
-                    Ok((key, value))
+                    let key = bytes_key.remove_column_variable(column).ok_or_else(|| {
+                        HotColdDBError::IterationError {
+                            unexpected_key: bytes_key.clone(),
+                        }
+                    })?;
+                    Ok((K::from_bytes(key)?, value))
                 }),
         )
     }
@@ -254,12 +250,12 @@ impl<E: EthSpec> KeyValueStore<E> for LevelDB<E> {
 impl<E: EthSpec> ItemStore<E> for LevelDB<E> {}
 
 /// Used for keying leveldb.
-#[derive(Debug, PartialEq)]
+#[derive(Debug, Clone, PartialEq, Eq, PartialOrd, Ord)]
 pub struct BytesKey {
     key: Vec<u8>,
 }
 
-impl Key for BytesKey {
+impl db_key::Key for BytesKey {
     fn from_u8(key: &[u8]) -> Self {
         Self { key: key.to_vec() }
     }
@@ -275,12 +271,20 @@ impl BytesKey {
         self.key.starts_with(column.as_bytes())
     }
 
-    /// Remove the column from a key, returning its `Hash256` portion.
+    /// Remove the column from a 32 byte key, yielding the `Hash256` key.
     pub fn remove_column(&self, column: DBColumn) -> Option<Hash256> {
+        let key = self.remove_column_variable(column)?;
+        (column.key_size() == 32).then(|| Hash256::from_slice(key))
+    }
+
+    /// Remove the column from a key.
+    ///
+    /// Will return `None` if the value doesn't match the column or has the wrong length.
+    pub fn remove_column_variable(&self, column: DBColumn) -> Option<&[u8]> {
         if self.matches_column(column) {
             let subkey = &self.key[column.as_bytes().len()..];
-            if subkey.len() == 32 {
-                return Some(Hash256::from_slice(subkey));
+            if subkey.len() == column.key_size() {
+                return Some(subkey);
             }
         }
         None

beacon_node/store/src/lib.rs

@@ -44,7 +44,7 @@ use std::sync::Arc;
 use strum::{EnumString, IntoStaticStr};
 pub use types::*;
 
-pub type ColumnIter<'a> = Box<dyn Iterator<Item = Result<(Hash256, Vec<u8>), Error>> + 'a>;
+pub type ColumnIter<'a, K> = Box<dyn Iterator<Item = Result<(K, Vec<u8>), Error>> + 'a>;
 pub type ColumnKeyIter<'a> = Box<dyn Iterator<Item = Result<Hash256, Error>> + 'a>;
 
 pub type RawEntryIter<'a> = Box<dyn Iterator<Item = Result<(Vec<u8>, Vec<u8>), Error>> + 'a>;
@@ -84,11 +84,12 @@ pub trait KeyValueStore<E: EthSpec>: Sync + Send + Sized + 'static {
     fn compact(&self) -> Result<(), Error>;
 
     /// Iterate through all keys and values in a particular column.
-    fn iter_column(&self, _column: DBColumn) -> ColumnIter {
-        // Default impl for non LevelDB databases
-        Box::new(std::iter::empty())
+    fn iter_column<K: Key>(&self, column: DBColumn) -> ColumnIter<K> {
+        self.iter_column_from(column, &vec![0; column.key_size()])
     }
 
+    fn iter_column_from<K: Key>(&self, column: DBColumn, from: &[u8]) -> ColumnIter<K>;
+
     fn iter_raw_entries(&self, _column: DBColumn, _prefix: &[u8]) -> RawEntryIter {
         Box::new(std::iter::empty())
     }
@@ -98,9 +99,26 @@ pub trait KeyValueStore<E: EthSpec>: Sync + Send + Sized + 'static {
     }
 
     /// Iterate through all keys in a particular column.
-    fn iter_column_keys(&self, _column: DBColumn) -> ColumnKeyIter {
-        // Default impl for non LevelDB databases
-        Box::new(std::iter::empty())
+    fn iter_column_keys(&self, column: DBColumn) -> ColumnKeyIter;
+}
+
+pub trait Key: Sized + 'static {
+    fn from_bytes(key: &[u8]) -> Result<Self, Error>;
+}
+
+impl Key for Hash256 {
+    fn from_bytes(key: &[u8]) -> Result<Self, Error> {
+        if key.len() == 32 {
+            Ok(Hash256::from_slice(key))
+        } else {
+            Err(Error::InvalidKey)
+        }
+    }
+}
+
+impl Key for Vec<u8> {
+    fn from_bytes(key: &[u8]) -> Result<Self, Error> {
+        Ok(key.to_vec())
     }
 }
 
@@ -250,6 +268,35 @@ impl DBColumn {
     pub fn as_bytes(self) -> &'static [u8] {
         self.as_str().as_bytes()
     }
+
+    /// Most database keys are 32 bytes, but some freezer DB keys are 8 bytes.
+    ///
+    /// This function returns the number of bytes used by keys in a given column.
+    pub fn key_size(self) -> usize {
+        match self {
+            Self::OverflowLRUCache => 40,
+            Self::BeaconMeta
+            | Self::BeaconBlock
+            | Self::BeaconState
+            | Self::BeaconBlob
+            | Self::BeaconStateSummary
+            | Self::BeaconStateTemporary
+            | Self::ExecPayload
+            | Self::BeaconChain
+            | Self::OpPool
+            | Self::Eth1Cache
+            | Self::ForkChoice
+            | Self::PubkeyCache
+            | Self::BeaconRestorePoint
+            | Self::DhtEnrs
+            | Self::OptimisticTransitionBlock => 32,
+            Self::BeaconBlockRoots
+            | Self::BeaconStateRoots
+            | Self::BeaconHistoricalRoots
+            | Self::BeaconHistoricalSummaries
+            | Self::BeaconRandaoMixes => 8,
+        }
+    }
 }
 
 /// An item that may stored in a `Store` by serializing and deserializing from bytes.

beacon_node/store/src/memory_store.rs

@@ -1,17 +1,17 @@
-use super::{Error, ItemStore, KeyValueStore, KeyValueStoreOp};
-use crate::{ColumnIter, DBColumn};
+use crate::{
+    get_key_for_col, leveldb_store::BytesKey, ColumnIter, ColumnKeyIter, DBColumn, Error,
+    ItemStore, Key, KeyValueStore, KeyValueStoreOp,
+};
 use parking_lot::{Mutex, MutexGuard, RwLock};
-use std::collections::{HashMap, HashSet};
+use std::collections::BTreeMap;
 use std::marker::PhantomData;
 use types::*;
 
-type DBHashMap = HashMap<Vec<u8>, Vec<u8>>;
-type DBKeyMap = HashMap<Vec<u8>, HashSet<Vec<u8>>>;
+type DBMap = BTreeMap<BytesKey, Vec<u8>>;
 
-/// A thread-safe `HashMap` wrapper.
+/// A thread-safe `BTreeMap` wrapper.
 pub struct MemoryStore<E: EthSpec> {
-    db: RwLock<DBHashMap>,
-    col_keys: RwLock<DBKeyMap>,
+    db: RwLock<DBMap>,
     transaction_mutex: Mutex<()>,
     _phantom: PhantomData<E>,
 }
@@ -20,36 +20,24 @@ impl<E: EthSpec> MemoryStore<E> {
     /// Create a new, empty database.
     pub fn open() -> Self {
         Self {
-            db: RwLock::new(HashMap::new()),
-            col_keys: RwLock::new(HashMap::new()),
+            db: RwLock::new(BTreeMap::new()),
             transaction_mutex: Mutex::new(()),
             _phantom: PhantomData,
         }
     }
-
-    fn get_key_for_col(col: &str, key: &[u8]) -> Vec<u8> {
-        let mut col = col.as_bytes().to_vec();
-        col.append(&mut key.to_vec());
-        col
-    }
 }
 
 impl<E: EthSpec> KeyValueStore<E> for MemoryStore<E> {
     /// Get the value of some key from the database. Returns `None` if the key does not exist.
     fn get_bytes(&self, col: &str, key: &[u8]) -> Result<Option<Vec<u8>>, Error> {
-        let column_key = Self::get_key_for_col(col, key);
+        let column_key = BytesKey::from_vec(get_key_for_col(col, key));
         Ok(self.db.read().get(&column_key).cloned())
     }
 
     /// Puts a key in the database.
     fn put_bytes(&self, col: &str, key: &[u8], val: &[u8]) -> Result<(), Error> {
-        let column_key = Self::get_key_for_col(col, key);
+        let column_key = BytesKey::from_vec(get_key_for_col(col, key));
         self.db.write().insert(column_key, val.to_vec());
-        self.col_keys
-            .write()
-            .entry(col.as_bytes().to_vec())
-            .or_default()
-            .insert(key.to_vec());
         Ok(())
     }
 
@@ -64,18 +52,14 @@ impl<E: EthSpec> KeyValueStore<E> for MemoryStore<E> {
 
     /// Return true if some key exists in some column.
     fn key_exists(&self, col: &str, key: &[u8]) -> Result<bool, Error> {
-        let column_key = Self::get_key_for_col(col, key);
+        let column_key = BytesKey::from_vec(get_key_for_col(col, key));
         Ok(self.db.read().contains_key(&column_key))
     }
 
     /// Delete some key from the database.
     fn key_delete(&self, col: &str, key: &[u8]) -> Result<(), Error> {
-        let column_key = Self::get_key_for_col(col, key);
+        let column_key = BytesKey::from_vec(get_key_for_col(col, key));
         self.db.write().remove(&column_key);
-        self.col_keys
-            .write()
-            .get_mut(&col.as_bytes().to_vec())
-            .map(|set| set.remove(key));
         Ok(())
     }
 
@@ -83,35 +67,41 @@ impl<E: EthSpec> KeyValueStore<E> for MemoryStore<E> {
         for op in batch {
             match op {
                 KeyValueStoreOp::PutKeyValue(key, value) => {
-                    self.db.write().insert(key, value);
+                    self.db.write().insert(BytesKey::from_vec(key), value);
                 }
 
-                KeyValueStoreOp::DeleteKey(hash) => {
-                    self.db.write().remove(&hash);
+                KeyValueStoreOp::DeleteKey(key) => {
+                    self.db.write().remove(&BytesKey::from_vec(key));
                 }
             }
         }
         Ok(())
     }
 
-    // pub type ColumnIter<'a> = Box<dyn Iterator<Item = Result<(Hash256, Vec<u8>), Error>> + 'a>;
-    fn iter_column(&self, column: DBColumn) -> ColumnIter {
+    fn iter_column_from<K: Key>(&self, column: DBColumn, from: &[u8]) -> ColumnIter<K> {
+        // We use this awkward pattern because we can't lock the `self.db` field *and* maintain a
+        // reference to the lock guard across calls to `.next()`. This would be require a
+        // struct with a field (the iterator) which references another field (the lock guard).
+        let start_key = BytesKey::from_vec(get_key_for_col(column.as_str(), from));
         let col = column.as_str();
-        if let Some(keys) = self
-            .col_keys
+        let keys = self
+            .db
             .read()
-            .get(col.as_bytes())
-            .map(|set| set.iter().cloned().collect::<Vec<_>>())
-        {
-            Box::new(keys.into_iter().filter_map(move |key| {
-                let hash = Hash256::from_slice(&key);
-                self.get_bytes(col, &key)
-                    .transpose()
-                    .map(|res| res.map(|bytes| (hash, bytes)))
-            }))
-        } else {
-            Box::new(std::iter::empty())
-        }
+            .range(start_key..)
+            .take_while(|(k, _)| k.remove_column_variable(column).is_some())
+            .filter_map(|(k, _)| k.remove_column_variable(column).map(|k| k.to_vec()))
+            .collect::<Vec<_>>();
+        Box::new(keys.into_iter().filter_map(move |key| {
+            self.get_bytes(col, &key).transpose().map(|res| {
+                let k = K::from_bytes(&key)?;
+                let v = res?;
+                Ok((k, v))
+            })
+        }))
+    }
+
+    fn iter_column_keys(&self, column: DBColumn) -> ColumnKeyIter {
+        Box::new(self.iter_column(column).map(|res| res.map(|(k, _)| k)))
     }
 
     fn begin_rw_transaction(&self) -> MutexGuard<()> {