Ethereum Block Storage

[Jan-Jan]

SCC lead: @pgherveou
Design doc: Ethereum Block Storage

Scope

• Review design doc
• Review whether or not the storage needs to be stored, or if we can just merkle root
• Add the new storage to pallet_revive
• Add the on_finalize hook to compute the merkle roots and block hash.
• Update pallet_revive blockhash host function to read from the new storage instead of frame_system hash
• Add benchmarks for on_finalize hook & for the extra cost incurred by each write in the eth_call / eth_instantiate dispatchable
• If needed add additional limit to ensure it can't be exploited to generate OOM errors in the PVF
• Adapt eth-rpc to take advantage of the new storage instead of re-building them dynamically from the block

[lexnv]

The workload is distributed between the following PRs:

• frame/revive: ETH block storage polkadot-sdk#9418
  • Add the main functionality of building the ETH block and hash
  • Must ensure that ETH_* calls are denied if called directly and propagate that info through the call stack (frame/revive: ETH block storage polkadot-sdk#9418 (comment))
• Add comprehensive test data for Ethereum trie root validation polkadot-sdk#9452
  • Improves the trie root and adds several tests, plus example binary to fetch the test data from real-world nodes
• Pallet revive benchmarks polkadot-sdk#9496
  • Captures the performance and weights of the pallet revive
• revive/rpc: Adjust the RPC to utilize the ETH block from storage polkadot-sdk#9512
  • Adjust the RPC layer to use the ETH block from storage

Followups

The following issues were identified while working on the main task; they are not blocking and can be addressed with a lower priority (sorted by most important):

• HashesOrTransactionInfos (de)serialization issue #146
• tests/flaky: Revive RPC runtime_api_dry_run_addr_works test is flaky polkadot-sdk#9528
• eth/revive-rpc: Use ChainHead rpc methods instead of legacy ones polkadot-sdk#9327

[lexnv]

Outstanding PRs with the trie of dependency:

• frame/revive: ETH block storage polkadot-sdk#9418
• Implements the main functionality
  • revive/types: Implement the ethereum 7702 transaction polkadot-sdk#9538
  • Fixes the 7702 transaction decoding and adds tests)
    • Compare incremental Trie Builder for space saveup optimizations polkadot-sdk#9554
    • Uses incremental trie builders to reduce mem usage)
  • Pallet revive benchmarks polkadot-sdk#9496
  • The benchmark work
  • revive/rpc: Adjust the RPC to utilize the ETH block from storage polkadot-sdk#9512
  • Initial RPC adjustments and approach
    • revive/rpc: add ethereum-substrate block hash mapping infrastructure polkadot-sdk#9616
    • Mappings of eth block hashes into substrate block hashes
  • Add comprehensive test data for Ethereum trie root validation polkadot-sdk#9452
  • Check trie computation and indepth testing

Review Notes

We'd need to determine first if the incremental approach is good enough in terms of mem consumption / tradeoffs:

• Compare incremental Trie Builder for space saveup optimizations polkadot-sdk#9554
  • the PR outlines the implementation details of the trie builders
  • the first transaction is temporary stored until tx number 128 (or the end)
  • although we are keeping the minimum data around, events could grow unboundedly

[adamant-al]

👍

[lrubasze]

Issue discovered during the block hash mapping implementation

• revive/rpc: serde limitation with flatten + untagged enums requires custom deserializer polkadot-sdk#9933