CAP: 0046-02 (formerly 0047)
Title: Smart Contract Lifecycle
Working Group:
Owner: Siddharth Suresh <@sisuresh>
Authors: Siddharth Suresh <@sisuresh>, Dmytro Kozhevin <@dmkozh>
Consulted: Graydon Hoare <@graydon>, Jon Jove <@jonjove>, Leigh McCulloch <@leighmcculloch>, Nicolas Barry <@MonsieurNicolas>
Status: Draft
Created: 2022-05-02
Discussion:
Protocol version: TBD
This proposal defines the structure of smart contracts on Stellar and specifies how users can create them.
See the Soroban overview CAP.
Users need a way to manage smart contracts on the network. This CAP allows users to deploy the smart contracts to the network and specifies the supported contract code kinds.
This CAP also lets the validators turn off all smart contract functionality if some unexpected behavior is found in the protocol.
See the XDR diffs in the Soroban overview CAP, specifically those referring to
HOST_FUNCTION_TYPE_CREATE_CONTRACT
and
HOST_FUNCTION_TYPE_INSTALL_CONTRACT_CODE
.
This defines the terms we use in the following sections without going into their design and implementation details.
Contract source can be thought of as a 'class' of a contract. Multiple contracts can share the same source, but have their own state. Thanks to that sharing capability, we can reduce the amount of duplication in ledger and only store unique contract sources.
This CAP defines two possible kinds of contract sources:
- WASM source: a blob of WASM code that is stored in a separate ledger entry and is deduplicated based on contents. This is installed to ledger by the users.
- Built-in contract: this is a 'source' compiled into host directly that has a protocol-defined interface and behavior.
Contract instance can be thought of as an instance of the contract 'class'. Contract instance consists of:
- Identifier: SHA-256 hash of a pre-image payload
- Source reference: a pointer to the WASM source or a tag of a built-in contract
A contract instance may own an arbitrary amount of ledger entries attributed to its identifier. Contracts that share the same source in no way may influence each other; from the perspective of a contract invoker there is no difference between calling the contracts with the same or different source references (besides the possible contract-defined behavior differences).
This CAP defines the following supported contract identifier preimage kinds. The
use cases for every identifier type are described in the following sections.
Every preimage is a part of the HashIDPreimage
union and has a unique tag
associated with it in order to ensure that there are no collisions with other
hashes in the protocol.
ENVELOPE_TYPE_CONTRACT_ID_FROM_ED25519
: built from aned25519
public key and the user-specifieduint256
salt.ENVELOPE_TYPE_CONTRACT_ID_FROM_SOURCE_ACCOUNT
: built from a Stellar account identifier and the user-specifieduint256
salt.ENVELOPE_TYPE_CONTRACT_ID_FROM_ASSET
: built from a StellarAsset
structure.ENVELOPE_TYPE_CONTRACT_ID_FROM_CONTRACT
: built from another contract identifier and contract-specifieduint256
salt.
Every preimage must also include a networkID
(a hash of the network
passphrase) which ensures that every network has unique set of contract
identifiers.
WASM contract sources can be installed to the network without instantiating a
contract via InvokeHostFunctionOp
(defined in CAP-0046-04)
with HOST_FUNCTION_TYPE_INSTALL_CONTRACT_CODE
host function type.
This function accepts InstallContractCodeArgs
struct that contains the WASM
contract code.
Installed contracts are stored in ContractCodeEntry
ledger entries. These
entries are keyed by the hash of InstallContractCodeArgs
used to install
them.
Contract installation host function will compute the hash of
InstallContractCodeArgs
and check if such a contract code already exists. If
the entry exists, the operation will immediately succeed. If it doesn't, the
new ContractCodeEntry
will be created.
Host does not perform any validation on the installed contract code, besides checking its size.
The maximum WASM contract size will be introduced as a ConfigSettingEntry
(see
CAP-0046-09 for details on config entries).
It is set during the protocol version upgrade using a new ConfigSettingEntry
,
with configSettingID
== CONFIG_SETTING_CONTRACT_MAX_SIZE_BYTES
, and
contractMaxSizeBytes
== 16384
. The valid values for
contractMaxSizeBytes
are[0, 256000] (inclusive).
This CAP does not specify a way to install WASM sources from within a contract.
This is done to encourage efficient code reuse and deduplication: if the contract was allowed to install the WASM code, then we'd need to store it twice (in the installer contract and in the source entry).
Contracts can be instantiated via InvokeHostFunctionOp
with
HOST_FUNCTION_TYPE_CREATE_CONTRACT
host function type.
The function accepts CreateContractArgs
struct that defines the input for
building the contract identifier preimage (contractID
field) and the contract
source reference (source
field).
All the preimage types besides ENVELOPE_TYPE_CONTRACT_ID_FROM_CONTRACT
can be
built from the contractID
field.
The source and identifier arguments are normally independent of each other with
an exception: identifiers that are built from CONTRACT_ID_FROM_ASSET
may only
be used in conjunction with built-in token contract source. This handles the
special case of instantiating token contracts corresponding to the classic
Stellar assets (see more details in CAP-0046-06).
The host builds the actual contract identifier by computing SHA-256 of the
HashIDPreimage
corresponding to the contractID
. If the contract identifier
already exists, the operation fails.
If the identifier is new, the host will a new ContractDataEntry
from
CAP-0046-05 with a SCV_STATIC
key type, and
SCS_LEDGER_KEY_CONTRACT_CODE
key value. The value of the entry is
SCContractCode
that either refers to the WASM code entry or to a built-in
contract (according to the value of source
field in CreateContractArgs
).
Building a ENVELOPE_TYPE_CONTRACT_ID_FROM_ED25519
preimage based on a public
ED25519 key has an a additional signature verification step as to make sure
that this key has authorized creating a contract on their behalf.
The owner of the key must sign SHA-256 hash of HashIDPreimage
of type
ENVELOPE_TYPE_CREATE_CONTRACT_ARGS
, that includes the network id, salt
, and
CreateContractSource
args that must match the respective args of
HOST_FUNCTION_TYPE_CREATE_CONTRACT
invocation.
One of the possible types of CreateContractSource
is
CONTRACT_SOURCE_INSTALLED
, that accepts InstallContractCodeArgs
. This is a
convenience argument that allows to install the code and instantiate a contract
using that code in a single operation.
The installation implementation is exactly the same as for the case when
HOST_FUNCTION_TYPE_INSTALL_CONTRACT_CODE
function is called. The contract
will be instantiated with WASM code reference source type that points to the
hash of the newly installed contract.
As mentioned in the installation section, if the contract code already exists in the ledger, the operation will still succeed, but no code entry will be created.
Factory contracts are quite popular already on other networks, so this CAP adds functionality to support them.
The following host functions are provided to instantiate contracts:
// Instantiates a contract with the source referring to the provided wasm_hash.
fn create_contract_from_contract(wasm_hash: Object /* 32-bytes array */,
salt: Object /* 32-bytes array */) -> Object /* 32-bytes array */
// Instantiates a contract with the source referring to the built-in token.
fn create_token_from_contract(salt: Object /* 32-bytes array */) -> Object /* 32-bytes array */
All of these functions return the identifier of the newly created contract.
The identifier of the created contract is generated by hashing the
HashIDPreimage
with type ENVELOPE_TYPE_CONTRACT_ID_FROM_CONTRACT
with the
salt provided by the host function call.
This proposal adds two new LedgerHeader
flags that can disable the create and
invoke contract operations using upgrades. The validators can use this
mechanism in case unexpected behaviour is seen. We also considered adding a
mechanism for validators to opt accounts into smart contracts to allow for
a "soft" launch, but the implementation changes to get this to work are not
simple. The validators have the LedgerHeader
overrides to fall back on, so
it's not clear that the complexity of adding a "soft" launch mechanism is worth
it.
Controls like pausing invocation or mutability for all or a subset of a contract should be put into a contract itself. Leaving it to the contract writer is a much more general solution than baking it into the protocol. The downside is this is more error prone and will take more space since the same logic will be implemented multiple times.
The entity that creates the ContractDataEntry
that contains the contract code
is not tied to it in any way. This allows for contract management and
authorization to be handled in the contract using whichever custom mechanism
the contract creator chooses.
Contract source code entries with the WASM code don't have any ownership. Anyone can install contract sources to the ledger and then anyone can use them. This encourages sharing the contract code and allows contracts that use it to be sure that their implementation can't unexpectedly change.
The contract code reference is stored in a ContractDataEntry
, but contract
code cannot be updated or deleted in the initial version. The host functions in
CAP-0046-05 to update or delete
ContractDataEntry
should trap if they are used on contract code.
The validators do not have a mechanism to ban specific contracts. Any kind of targeted banning mechanism can be worked around quite easily by creating new accounts and contracts.
The maximum contract size will be set during the protocol upgrade, and can be updated by the validators. This allows to adjust the contract sizes depending on the demand and network load requirements.
Pulling contractIDs from LedgerHeader.idPool
would be easier but it would make
parallelizing contract creation more difficult in the future. It's also more
difficult to determine what the contractID will be since the id pool would be
used by offers and other contracts. This CAP uses a Hash
instead as the
contractID.
With this CAP we provide several ways of building the contractID preimages that can be reproduced off-chain and then used to address the contracts that may or may not exist (for example, some general contracts like tokens or AMMs).
The security concerns from CAP-0046 (https://github.com/stellar/stellar-protocol/blob/master/core/cap-0046.md#security-concerns) apply here as well.
In addition to those concerns, this CAP does not provide validators with much control over contracts on the network. The only mechanism they have is blocking all contract creations and invocations, which should only be used in drastic situations. This CAP does not define a way for validators to block specific contracts.
While the contracts are immutable in this CAP, it's already possible to make them 'mutable' via proxy contracts (for example, contract A forwards its method calls to contract B and ID of contract B is stored in the data of the contract A).
To further support mutation via the proxy pattern we could do the following:
- Allow contract to modify its
SCS_LEDGER_KEY_CONTRACT_CODE
entry (as the modification can only happen from within the contract, this would need to be implemented in the first installed version of the contract) - Allow specifying a contract ID as the contract's source reference, so that the contract would be guaranteed to have exactly the same implementation as the referred contract without any additional code.