EIP-7377: Migration Transaction
Allow EOAs to send a one-time transaction which deploys code at their account.
Abstract
Introduce a new EIP-2718 transaction type with the format 0x04 || rlp([chainId, nonce, maxFeePerGas, maxPriorityFeePerGas, gasLimit, codeAddr, storage, data, value, accessList, yParity, r, s])
which sets the sending account's code
field in the state trie to the code
value at codeAddr
and applies the storage tuples to the sender's storage trie.
Motivation
Smart contract wallets have long been touted as the solution to Ethereum's user experience woes. As early as 2015, there were proposals for allowing smart contracts to originate transactions in hopes that new users would flock to smart contract wallets to store their assets. So far, only a fraction of users have elected to do so.
Today, account abstraction is still an important goal in Ethereum and there are many efforts attempting to realize it. We're getting closer to succeeding at this, but unfortunately the years of failure have caused many users to simply rely on EOA.
After a user has accumulated enough assets in an EOA, it is not tenable to migrate each individual asset to a new address. This is due both to the cost and to needing to manually sign and verify potentially hundreds of transactions.
This is an overlooked piece of the problem. Converting existing users to smart contract wallets efficiently will expedite adoption and push forward better support and integrations for smart contract wallets. They will no longer be dismissed as a niche use case.
Therefore, we must provide a mechanism, embedded in the protocol, to migrate EOAs to smart contracts. This EIP proposes such mechanism.
Specification
At the fork block X
, introduce the migration transaction type.
Migration Transaction
Definition
field | type |
---|---|
chainId | uint256 |
nonce | uint64 |
maxFeePerGas | uint256 |
maxPriorityFeePerGas | uint256 |
gasLimit | uint64 |
codeAddr | address |
storage | List[Tuple[uint256, uint256]] |
data | bytes |
value | uint256 |
accessList | List[Tuple[address, List[uint256]]] |
yParity | uint8 |
r | uint256 |
s | uint256 |
The EIP-2718 TransactionType
is 0x04
and the TransactionPayload
is rlp([chainId, nonce, maxFeePerGas, maxPriorityFeePerGas, gasLimit, codeAddr, storage, data, value, accessList, yParity, r, s])
.
The transaction's signature hash is keccak256(0x04 || rlp([chainId, nonce, maxFeePerGas, maxPriorityFeePerGas, gasLimit, codeAddr, storage, data, value, accessList])
Validation
A migration transaction is considered valid if the follow properties hold:
- all EIP-1559 properties, unless specified otherwise
- the code at
codeAddr
is less than the EIP-170 limit of24576
- the code at
codeAddr
must not have size0
The intrinsic gas calculation modified from EIP-1559 to be 21000 + 16 * non-zero calldata bytes + 4 * zero calldata bytes + 1900 * access list storage key count + 2400 * access list address count + 20000 * length of storage
.
Processing
Executing a migration transaction has two parts.
Contract Deployment
Unlike standard contract deployment, a migration transaction directly specifies what code
value the sender's account should be set to.
As the first step of processing the transaction, set the sender's code
to state[tx.codeAddr].code
. Next, for each tuple in tx.storage
and the sender's storage trie, set storage[t.first] = t.second
.
Transaction Execution
Now instantiate an EVM call into the sender's account using the same rules as EIP-1559 and set the transaction's origin to be keccak256(sender)[0..20]
.
Rationale
No to
address field
This transaction is only good for one-time use to migrate an EOA to a smart contract. It is designed to immediately call the deployed contract, which is at the sender's address, after deployment to allow the sender to do any kind of further processing.
Code pointer for deployment
Naively, one could design the migration transaction to have a field code
of type bytes
. However, there would be substantial duplication of code calldata, since many users will want to deploy the exact same thing (often a wallet). Using a pointer instead acknowledges this overwhelming use case for the transaction type, and exploits it as an optimization.
Cheaper storage
Since the storage is guaranteed to be empty, there is no need to read before write. This means only 20,000 gas is needed to pay for the EIP-2200 SSTORE_SET_GAS
value. This is a small discount to the normal cost of 22,100
, which is SSTORE_SET_GAS
plus the EIP-2929 COLD_SLOAD_COST
of 2100
, because no load occurs.
Intrinsic does not account for contract deployment
This takes advantage of the fact that clients tend to store a single, unique copy of code; no matter the number of deployments. Therefore, the only operation here is changing a pointer in the state trie to the desired code.
Additionally, the EOA already exists because it has enough balance for the migration transaction to be considered valid. Therefore, we don't need to pay a premium for adding a new account into the state trie.
Manipulating transaction origin
Many applications have a security check caller == origin
to verify the caller is an EOA. This is done to "protect" assets. While it is usually more of a bandage than an actual fix, we attempt to placate these projects by modifying the origin of the transaction so the check will continue performing its duty.
One-time migration
There is no technical reason we couldn't allow EOAs to change their code at any time with this transaction type. The only inhibitor at the moment is EIP-3607 which will cause migration transactions to be considered invalid if they come from an account with code already deployed. A functional reason for retaining this behavior though is that it makes it simpler to reason about contracts and their upgradability.
Backwards Compatibility
No backward compatibility issues found.
Security Considerations
Blind Signing
As with all sufficiently sophisticated account designs, if a user can be convinced to sign an arbitrary message, that message could be a migration transaction which is owned by a malicious actor instead of the user. This can generally be avoided if wallets treat these transactions with extreme care and create as much friction and verification as possible before completing the signature.
On ecrecover
Applications standards such as ERC-2612: Permit Extension have exploited the cryptographic relationship between EOA addresses and their private keys. Many tokens today support this extension, allowing EOAs to approve the transfer of fund from their account using only a signature. Although collisions between EOAs and contract accounts are considered unlikely and maybe impossible given today's computing power, this EIP would make it common place for private keys to exist for contract accounts. There are some considerations here regarding security:
- The obvious attack is a defi protocol deploys some their contract using this EIP and later sign an ERC-2612 message to steal the funds accrued in the contract. This can be avoided by wallets simply not allowing users to interact with protocols deployed in this manner.
- It's also worth mentioning that there are concerns around how this EIP will affect the cross chain experience. Ultimately a users private key may still have some control over the account's assets, depending on the exact protocols used on Ethereum and on other chains. It isn't really possible perfectly migrate the EOA at the same time, on all chains. The best thing that can be done is to educate the user that just because their account has been migrated doesn't mean that they are safe to now publicly reveal their private key. This seems like a reasonable request, especially since they'll want to retain the private key in case they want to use the address on any other EVM-like chain.
Something that may alleviate these issues to some degree would be to add an EXTCODEHASH
check in ecrecover
. If the recovered account has code, the precompile will revert. This would disallow migrated EOAs from using standards like ERC-2612.
Copyright
Copyright and related rights waived via CC0.