serai/processor/ethereum/router/contracts/Router.sol

241 lines
9.1 KiB
Solidity
Raw Normal View History

// SPDX-License-Identifier: AGPL-3.0-only
pragma solidity ^0.8.26;
2024-09-17 05:04:08 +00:00
import "IERC20.sol";
import "Schnorr.sol";
// _ is used as a prefix for internal functions and smart-contract-scoped variables
Ethereum Integration (#557) * Clean up Ethereum * Consistent contract address for deployed contracts * Flesh out Router a bit * Add a Deployer for DoS-less deployment * Implement Router-finding * Use CREATE2 helper present in ethers * Move from CREATE2 to CREATE Bit more streamlined for our use case. * Document ethereum-serai * Tidy tests a bit * Test updateSeraiKey * Use encodePacked for updateSeraiKey * Take in the block hash to read state during * Add a Sandbox contract to the Ethereum integration * Add retrieval of transfers from Ethereum * Add inInstruction function to the Router * Augment our handling of InInstructions events with a check the transfer event also exists * Have the Deployer error upon failed deployments * Add --via-ir * Make get_transaction test-only We only used it to get transactions to confirm the resolution of Eventualities. Eventualities need to be modularized. By introducing the dedicated confirm_completion function, we remove the need for a non-test get_transaction AND begin this modularization (by no longer explicitly grabbing a transaction to check with). * Modularize Eventuality Almost fully-deprecates the Transaction trait for Completion. Replaces Transaction ID with Claim. * Modularize the Scheduler behind a trait * Add an extremely basic account Scheduler * Add nonce uses, key rotation to the account scheduler * Only report the account Scheduler empty after transferring keys Also ban payments to the branch/change/forward addresses. * Make fns reliant on state test-only * Start of an Ethereum integration for the processor * Add a session to the Router to prevent updateSeraiKey replaying This would only happen if an old key was rotated to again, which would require n-of-n collusion (already ridiculous and a valid fault attributable event). It just clarifies the formal arguments. * Add a RouterCommand + SignMachine for producing it to coins/ethereum * Ethereum which compiles * Have branch/change/forward return an option Also defines a UtxoNetwork extension trait for MAX_INPUTS. * Make external_address exclusively a test fn * Move the "account" scheduler to "smart contract" * Remove ABI artifact * Move refund/forward Plan creation into the Processor We create forward Plans in the scan path, and need to know their exact fees in the scan path. This requires adding a somewhat wonky shim_forward_plan method so we can obtain a Plan equivalent to the actual forward Plan for fee reasons, yet don't expect it to be the actual forward Plan (which may be distinct if the Plan pulls from the global state, such as with a nonce). Also properly types a Scheduler addendum such that the SC scheduler isn't cramming the nonce to use into the N::Output type. * Flesh out the Ethereum integration more * Two commits ago, into the **Scheduler, not Processor * Remove misc TODOs in SC Scheduler * Add constructor to RouterCommandMachine * RouterCommand read, pairing with the prior added write * Further add serialization methods * Have the Router's key included with the InInstruction This does not use the key at the time of the event. This uses the key at the end of the block for the event. Its much simpler than getting the full event streams for each, checking when they interlace. This does not read the state. Every block, this makes a request for every single key update and simply chooses the last one. This allows pruning state, only keeping the event tree. Ideally, we'd also introduce a cache to reduce the cost of the filter (small in events yielded, long in blocks searched). Since Serai doesn't have any forwarding TXs, nor Branches, nor change, all of our Plans should solely have payments out, and there's no expectation of a Plan being made under one key broken by it being received by another key. * Add read/write to InInstruction * Abstract the ABI for Call/OutInstruction in ethereum-serai * Fill out signable_transaction for Ethereum * Move ethereum-serai to alloy Resolves #331. * Use the opaque sol macro instead of generated files * Move the processor over to the now-alloy-based ethereum-serai * Use the ecrecover provided by alloy * Have the SC use nonce for rotation, not session (an independent nonce which wasn't synchronized) * Always use the latest keys for SC scheduled plans * get_eventuality_completions for Ethereum * Finish fleshing out the processor Ethereum integration as needed for serai-processor tests This doesn't not support any actual deployments, not even the ones simulated by serai-processor-docker-tests. * Add alloy-simple-request-transport to the GH workflows * cargo update * Clarify a few comments and make one check more robust * Use a string for 27.0 in .github * Remove optional from no-longer-optional dependencies in processor * Add alloy to git deny exception * Fix no longer optional specification in processor's binaries feature * Use a version of foundry from 2024 * Correct fetching Bitcoin TXs in the processor docker tests * Update rustls to resolve RUSTSEC warnings * Use the monthly nightly foundry, not the deleted daily nightly
2024-04-21 10:02:12 +00:00
contract Router {
// Nonce is incremented for each command executed, preventing replays
uint256 private _nonce;
// The nonce which will be used for the smart contracts we deploy, enabling
// predicting their addresses
uint256 private _smartContractNonce;
// The current public key, defined as per the Schnorr library
bytes32 private _seraiKey;
enum DestinationType {
Address,
Code
}
Ethereum Integration (#557) * Clean up Ethereum * Consistent contract address for deployed contracts * Flesh out Router a bit * Add a Deployer for DoS-less deployment * Implement Router-finding * Use CREATE2 helper present in ethers * Move from CREATE2 to CREATE Bit more streamlined for our use case. * Document ethereum-serai * Tidy tests a bit * Test updateSeraiKey * Use encodePacked for updateSeraiKey * Take in the block hash to read state during * Add a Sandbox contract to the Ethereum integration * Add retrieval of transfers from Ethereum * Add inInstruction function to the Router * Augment our handling of InInstructions events with a check the transfer event also exists * Have the Deployer error upon failed deployments * Add --via-ir * Make get_transaction test-only We only used it to get transactions to confirm the resolution of Eventualities. Eventualities need to be modularized. By introducing the dedicated confirm_completion function, we remove the need for a non-test get_transaction AND begin this modularization (by no longer explicitly grabbing a transaction to check with). * Modularize Eventuality Almost fully-deprecates the Transaction trait for Completion. Replaces Transaction ID with Claim. * Modularize the Scheduler behind a trait * Add an extremely basic account Scheduler * Add nonce uses, key rotation to the account scheduler * Only report the account Scheduler empty after transferring keys Also ban payments to the branch/change/forward addresses. * Make fns reliant on state test-only * Start of an Ethereum integration for the processor * Add a session to the Router to prevent updateSeraiKey replaying This would only happen if an old key was rotated to again, which would require n-of-n collusion (already ridiculous and a valid fault attributable event). It just clarifies the formal arguments. * Add a RouterCommand + SignMachine for producing it to coins/ethereum * Ethereum which compiles * Have branch/change/forward return an option Also defines a UtxoNetwork extension trait for MAX_INPUTS. * Make external_address exclusively a test fn * Move the "account" scheduler to "smart contract" * Remove ABI artifact * Move refund/forward Plan creation into the Processor We create forward Plans in the scan path, and need to know their exact fees in the scan path. This requires adding a somewhat wonky shim_forward_plan method so we can obtain a Plan equivalent to the actual forward Plan for fee reasons, yet don't expect it to be the actual forward Plan (which may be distinct if the Plan pulls from the global state, such as with a nonce). Also properly types a Scheduler addendum such that the SC scheduler isn't cramming the nonce to use into the N::Output type. * Flesh out the Ethereum integration more * Two commits ago, into the **Scheduler, not Processor * Remove misc TODOs in SC Scheduler * Add constructor to RouterCommandMachine * RouterCommand read, pairing with the prior added write * Further add serialization methods * Have the Router's key included with the InInstruction This does not use the key at the time of the event. This uses the key at the end of the block for the event. Its much simpler than getting the full event streams for each, checking when they interlace. This does not read the state. Every block, this makes a request for every single key update and simply chooses the last one. This allows pruning state, only keeping the event tree. Ideally, we'd also introduce a cache to reduce the cost of the filter (small in events yielded, long in blocks searched). Since Serai doesn't have any forwarding TXs, nor Branches, nor change, all of our Plans should solely have payments out, and there's no expectation of a Plan being made under one key broken by it being received by another key. * Add read/write to InInstruction * Abstract the ABI for Call/OutInstruction in ethereum-serai * Fill out signable_transaction for Ethereum * Move ethereum-serai to alloy Resolves #331. * Use the opaque sol macro instead of generated files * Move the processor over to the now-alloy-based ethereum-serai * Use the ecrecover provided by alloy * Have the SC use nonce for rotation, not session (an independent nonce which wasn't synchronized) * Always use the latest keys for SC scheduled plans * get_eventuality_completions for Ethereum * Finish fleshing out the processor Ethereum integration as needed for serai-processor tests This doesn't not support any actual deployments, not even the ones simulated by serai-processor-docker-tests. * Add alloy-simple-request-transport to the GH workflows * cargo update * Clarify a few comments and make one check more robust * Use a string for 27.0 in .github * Remove optional from no-longer-optional dependencies in processor * Add alloy to git deny exception * Fix no longer optional specification in processor's binaries feature * Use a version of foundry from 2024 * Correct fetching Bitcoin TXs in the processor docker tests * Update rustls to resolve RUSTSEC warnings * Use the monthly nightly foundry, not the deleted daily nightly
2024-04-21 10:02:12 +00:00
2024-09-17 05:04:08 +00:00
struct AddressDestination {
address destination;
}
struct CodeDestination {
uint32 gas_limit;
2024-09-17 05:04:08 +00:00
bytes code;
}
struct OutInstruction {
DestinationType destinationType;
bytes destination;
uint256 value;
}
struct Signature {
bytes32 c;
bytes32 s;
}
event SeraiKeyUpdated(uint256 indexed nonce, bytes32 indexed key);
2024-09-17 01:34:59 +00:00
event InInstruction(
address indexed from, address indexed coin, uint256 amount, bytes instruction
);
2024-09-17 05:04:08 +00:00
event Executed(uint256 indexed nonce, bytes32 indexed message_hash);
Ethereum Integration (#557) * Clean up Ethereum * Consistent contract address for deployed contracts * Flesh out Router a bit * Add a Deployer for DoS-less deployment * Implement Router-finding * Use CREATE2 helper present in ethers * Move from CREATE2 to CREATE Bit more streamlined for our use case. * Document ethereum-serai * Tidy tests a bit * Test updateSeraiKey * Use encodePacked for updateSeraiKey * Take in the block hash to read state during * Add a Sandbox contract to the Ethereum integration * Add retrieval of transfers from Ethereum * Add inInstruction function to the Router * Augment our handling of InInstructions events with a check the transfer event also exists * Have the Deployer error upon failed deployments * Add --via-ir * Make get_transaction test-only We only used it to get transactions to confirm the resolution of Eventualities. Eventualities need to be modularized. By introducing the dedicated confirm_completion function, we remove the need for a non-test get_transaction AND begin this modularization (by no longer explicitly grabbing a transaction to check with). * Modularize Eventuality Almost fully-deprecates the Transaction trait for Completion. Replaces Transaction ID with Claim. * Modularize the Scheduler behind a trait * Add an extremely basic account Scheduler * Add nonce uses, key rotation to the account scheduler * Only report the account Scheduler empty after transferring keys Also ban payments to the branch/change/forward addresses. * Make fns reliant on state test-only * Start of an Ethereum integration for the processor * Add a session to the Router to prevent updateSeraiKey replaying This would only happen if an old key was rotated to again, which would require n-of-n collusion (already ridiculous and a valid fault attributable event). It just clarifies the formal arguments. * Add a RouterCommand + SignMachine for producing it to coins/ethereum * Ethereum which compiles * Have branch/change/forward return an option Also defines a UtxoNetwork extension trait for MAX_INPUTS. * Make external_address exclusively a test fn * Move the "account" scheduler to "smart contract" * Remove ABI artifact * Move refund/forward Plan creation into the Processor We create forward Plans in the scan path, and need to know their exact fees in the scan path. This requires adding a somewhat wonky shim_forward_plan method so we can obtain a Plan equivalent to the actual forward Plan for fee reasons, yet don't expect it to be the actual forward Plan (which may be distinct if the Plan pulls from the global state, such as with a nonce). Also properly types a Scheduler addendum such that the SC scheduler isn't cramming the nonce to use into the N::Output type. * Flesh out the Ethereum integration more * Two commits ago, into the **Scheduler, not Processor * Remove misc TODOs in SC Scheduler * Add constructor to RouterCommandMachine * RouterCommand read, pairing with the prior added write * Further add serialization methods * Have the Router's key included with the InInstruction This does not use the key at the time of the event. This uses the key at the end of the block for the event. Its much simpler than getting the full event streams for each, checking when they interlace. This does not read the state. Every block, this makes a request for every single key update and simply chooses the last one. This allows pruning state, only keeping the event tree. Ideally, we'd also introduce a cache to reduce the cost of the filter (small in events yielded, long in blocks searched). Since Serai doesn't have any forwarding TXs, nor Branches, nor change, all of our Plans should solely have payments out, and there's no expectation of a Plan being made under one key broken by it being received by another key. * Add read/write to InInstruction * Abstract the ABI for Call/OutInstruction in ethereum-serai * Fill out signable_transaction for Ethereum * Move ethereum-serai to alloy Resolves #331. * Use the opaque sol macro instead of generated files * Move the processor over to the now-alloy-based ethereum-serai * Use the ecrecover provided by alloy * Have the SC use nonce for rotation, not session (an independent nonce which wasn't synchronized) * Always use the latest keys for SC scheduled plans * get_eventuality_completions for Ethereum * Finish fleshing out the processor Ethereum integration as needed for serai-processor tests This doesn't not support any actual deployments, not even the ones simulated by serai-processor-docker-tests. * Add alloy-simple-request-transport to the GH workflows * cargo update * Clarify a few comments and make one check more robust * Use a string for 27.0 in .github * Remove optional from no-longer-optional dependencies in processor * Add alloy to git deny exception * Fix no longer optional specification in processor's binaries feature * Use a version of foundry from 2024 * Correct fetching Bitcoin TXs in the processor docker tests * Update rustls to resolve RUSTSEC warnings * Use the monthly nightly foundry, not the deleted daily nightly
2024-04-21 10:02:12 +00:00
error InvalidSignature();
error InvalidAmount();
error FailedTransfer();
// Update the Serai key at the end of the current function.
modifier _updateSeraiKeyAtEndOfFn(uint256 nonceUpdatedWith, bytes32 newSeraiKey) {
// Run the function itself.
Ethereum Integration (#557) * Clean up Ethereum * Consistent contract address for deployed contracts * Flesh out Router a bit * Add a Deployer for DoS-less deployment * Implement Router-finding * Use CREATE2 helper present in ethers * Move from CREATE2 to CREATE Bit more streamlined for our use case. * Document ethereum-serai * Tidy tests a bit * Test updateSeraiKey * Use encodePacked for updateSeraiKey * Take in the block hash to read state during * Add a Sandbox contract to the Ethereum integration * Add retrieval of transfers from Ethereum * Add inInstruction function to the Router * Augment our handling of InInstructions events with a check the transfer event also exists * Have the Deployer error upon failed deployments * Add --via-ir * Make get_transaction test-only We only used it to get transactions to confirm the resolution of Eventualities. Eventualities need to be modularized. By introducing the dedicated confirm_completion function, we remove the need for a non-test get_transaction AND begin this modularization (by no longer explicitly grabbing a transaction to check with). * Modularize Eventuality Almost fully-deprecates the Transaction trait for Completion. Replaces Transaction ID with Claim. * Modularize the Scheduler behind a trait * Add an extremely basic account Scheduler * Add nonce uses, key rotation to the account scheduler * Only report the account Scheduler empty after transferring keys Also ban payments to the branch/change/forward addresses. * Make fns reliant on state test-only * Start of an Ethereum integration for the processor * Add a session to the Router to prevent updateSeraiKey replaying This would only happen if an old key was rotated to again, which would require n-of-n collusion (already ridiculous and a valid fault attributable event). It just clarifies the formal arguments. * Add a RouterCommand + SignMachine for producing it to coins/ethereum * Ethereum which compiles * Have branch/change/forward return an option Also defines a UtxoNetwork extension trait for MAX_INPUTS. * Make external_address exclusively a test fn * Move the "account" scheduler to "smart contract" * Remove ABI artifact * Move refund/forward Plan creation into the Processor We create forward Plans in the scan path, and need to know their exact fees in the scan path. This requires adding a somewhat wonky shim_forward_plan method so we can obtain a Plan equivalent to the actual forward Plan for fee reasons, yet don't expect it to be the actual forward Plan (which may be distinct if the Plan pulls from the global state, such as with a nonce). Also properly types a Scheduler addendum such that the SC scheduler isn't cramming the nonce to use into the N::Output type. * Flesh out the Ethereum integration more * Two commits ago, into the **Scheduler, not Processor * Remove misc TODOs in SC Scheduler * Add constructor to RouterCommandMachine * RouterCommand read, pairing with the prior added write * Further add serialization methods * Have the Router's key included with the InInstruction This does not use the key at the time of the event. This uses the key at the end of the block for the event. Its much simpler than getting the full event streams for each, checking when they interlace. This does not read the state. Every block, this makes a request for every single key update and simply chooses the last one. This allows pruning state, only keeping the event tree. Ideally, we'd also introduce a cache to reduce the cost of the filter (small in events yielded, long in blocks searched). Since Serai doesn't have any forwarding TXs, nor Branches, nor change, all of our Plans should solely have payments out, and there's no expectation of a Plan being made under one key broken by it being received by another key. * Add read/write to InInstruction * Abstract the ABI for Call/OutInstruction in ethereum-serai * Fill out signable_transaction for Ethereum * Move ethereum-serai to alloy Resolves #331. * Use the opaque sol macro instead of generated files * Move the processor over to the now-alloy-based ethereum-serai * Use the ecrecover provided by alloy * Have the SC use nonce for rotation, not session (an independent nonce which wasn't synchronized) * Always use the latest keys for SC scheduled plans * get_eventuality_completions for Ethereum * Finish fleshing out the processor Ethereum integration as needed for serai-processor tests This doesn't not support any actual deployments, not even the ones simulated by serai-processor-docker-tests. * Add alloy-simple-request-transport to the GH workflows * cargo update * Clarify a few comments and make one check more robust * Use a string for 27.0 in .github * Remove optional from no-longer-optional dependencies in processor * Add alloy to git deny exception * Fix no longer optional specification in processor's binaries feature * Use a version of foundry from 2024 * Correct fetching Bitcoin TXs in the processor docker tests * Update rustls to resolve RUSTSEC warnings * Use the monthly nightly foundry, not the deleted daily nightly
2024-04-21 10:02:12 +00:00
_;
// Update the key.
_seraiKey = newSeraiKey;
emit SeraiKeyUpdated(nonceUpdatedWith, newSeraiKey);
}
constructor(bytes32 initialSeraiKey) _updateSeraiKeyAtEndOfFn(0, initialSeraiKey) {
// We consumed nonce 0 when setting the initial Serai key
_nonce = 1;
// Nonces are incremented by 1 upon account creation, prior to any code execution, per EIP-161
// This is incompatible with any networks which don't have their nonces start at 0
_smartContractNonce = 1;
}
// updateSeraiKey validates the given Schnorr signature against the current public key, and if
// successful, updates the contract's public key to the one specified.
2024-09-17 01:34:59 +00:00
function updateSeraiKey(bytes32 newSeraiKey, Signature calldata signature)
external
_updateSeraiKeyAtEndOfFn(_nonce, newSeraiKey)
{
// This DST needs a length prefix as well to prevent DSTs potentially being substrings of each
// other, yet this fine for our very well-defined, limited use
bytes32 message =
keccak256(abi.encodePacked("updateSeraiKey", block.chainid, _nonce, newSeraiKey));
_nonce++;
if (!Schnorr.verify(_seraiKey, message, signature.c, signature.s)) {
Ethereum Integration (#557) * Clean up Ethereum * Consistent contract address for deployed contracts * Flesh out Router a bit * Add a Deployer for DoS-less deployment * Implement Router-finding * Use CREATE2 helper present in ethers * Move from CREATE2 to CREATE Bit more streamlined for our use case. * Document ethereum-serai * Tidy tests a bit * Test updateSeraiKey * Use encodePacked for updateSeraiKey * Take in the block hash to read state during * Add a Sandbox contract to the Ethereum integration * Add retrieval of transfers from Ethereum * Add inInstruction function to the Router * Augment our handling of InInstructions events with a check the transfer event also exists * Have the Deployer error upon failed deployments * Add --via-ir * Make get_transaction test-only We only used it to get transactions to confirm the resolution of Eventualities. Eventualities need to be modularized. By introducing the dedicated confirm_completion function, we remove the need for a non-test get_transaction AND begin this modularization (by no longer explicitly grabbing a transaction to check with). * Modularize Eventuality Almost fully-deprecates the Transaction trait for Completion. Replaces Transaction ID with Claim. * Modularize the Scheduler behind a trait * Add an extremely basic account Scheduler * Add nonce uses, key rotation to the account scheduler * Only report the account Scheduler empty after transferring keys Also ban payments to the branch/change/forward addresses. * Make fns reliant on state test-only * Start of an Ethereum integration for the processor * Add a session to the Router to prevent updateSeraiKey replaying This would only happen if an old key was rotated to again, which would require n-of-n collusion (already ridiculous and a valid fault attributable event). It just clarifies the formal arguments. * Add a RouterCommand + SignMachine for producing it to coins/ethereum * Ethereum which compiles * Have branch/change/forward return an option Also defines a UtxoNetwork extension trait for MAX_INPUTS. * Make external_address exclusively a test fn * Move the "account" scheduler to "smart contract" * Remove ABI artifact * Move refund/forward Plan creation into the Processor We create forward Plans in the scan path, and need to know their exact fees in the scan path. This requires adding a somewhat wonky shim_forward_plan method so we can obtain a Plan equivalent to the actual forward Plan for fee reasons, yet don't expect it to be the actual forward Plan (which may be distinct if the Plan pulls from the global state, such as with a nonce). Also properly types a Scheduler addendum such that the SC scheduler isn't cramming the nonce to use into the N::Output type. * Flesh out the Ethereum integration more * Two commits ago, into the **Scheduler, not Processor * Remove misc TODOs in SC Scheduler * Add constructor to RouterCommandMachine * RouterCommand read, pairing with the prior added write * Further add serialization methods * Have the Router's key included with the InInstruction This does not use the key at the time of the event. This uses the key at the end of the block for the event. Its much simpler than getting the full event streams for each, checking when they interlace. This does not read the state. Every block, this makes a request for every single key update and simply chooses the last one. This allows pruning state, only keeping the event tree. Ideally, we'd also introduce a cache to reduce the cost of the filter (small in events yielded, long in blocks searched). Since Serai doesn't have any forwarding TXs, nor Branches, nor change, all of our Plans should solely have payments out, and there's no expectation of a Plan being made under one key broken by it being received by another key. * Add read/write to InInstruction * Abstract the ABI for Call/OutInstruction in ethereum-serai * Fill out signable_transaction for Ethereum * Move ethereum-serai to alloy Resolves #331. * Use the opaque sol macro instead of generated files * Move the processor over to the now-alloy-based ethereum-serai * Use the ecrecover provided by alloy * Have the SC use nonce for rotation, not session (an independent nonce which wasn't synchronized) * Always use the latest keys for SC scheduled plans * get_eventuality_completions for Ethereum * Finish fleshing out the processor Ethereum integration as needed for serai-processor tests This doesn't not support any actual deployments, not even the ones simulated by serai-processor-docker-tests. * Add alloy-simple-request-transport to the GH workflows * cargo update * Clarify a few comments and make one check more robust * Use a string for 27.0 in .github * Remove optional from no-longer-optional dependencies in processor * Add alloy to git deny exception * Fix no longer optional specification in processor's binaries feature * Use a version of foundry from 2024 * Correct fetching Bitcoin TXs in the processor docker tests * Update rustls to resolve RUSTSEC warnings * Use the monthly nightly foundry, not the deleted daily nightly
2024-04-21 10:02:12 +00:00
revert InvalidSignature();
}
}
2024-09-17 01:34:59 +00:00
function inInstruction(address coin, uint256 amount, bytes memory instruction) external payable {
Ethereum Integration (#557) * Clean up Ethereum * Consistent contract address for deployed contracts * Flesh out Router a bit * Add a Deployer for DoS-less deployment * Implement Router-finding * Use CREATE2 helper present in ethers * Move from CREATE2 to CREATE Bit more streamlined for our use case. * Document ethereum-serai * Tidy tests a bit * Test updateSeraiKey * Use encodePacked for updateSeraiKey * Take in the block hash to read state during * Add a Sandbox contract to the Ethereum integration * Add retrieval of transfers from Ethereum * Add inInstruction function to the Router * Augment our handling of InInstructions events with a check the transfer event also exists * Have the Deployer error upon failed deployments * Add --via-ir * Make get_transaction test-only We only used it to get transactions to confirm the resolution of Eventualities. Eventualities need to be modularized. By introducing the dedicated confirm_completion function, we remove the need for a non-test get_transaction AND begin this modularization (by no longer explicitly grabbing a transaction to check with). * Modularize Eventuality Almost fully-deprecates the Transaction trait for Completion. Replaces Transaction ID with Claim. * Modularize the Scheduler behind a trait * Add an extremely basic account Scheduler * Add nonce uses, key rotation to the account scheduler * Only report the account Scheduler empty after transferring keys Also ban payments to the branch/change/forward addresses. * Make fns reliant on state test-only * Start of an Ethereum integration for the processor * Add a session to the Router to prevent updateSeraiKey replaying This would only happen if an old key was rotated to again, which would require n-of-n collusion (already ridiculous and a valid fault attributable event). It just clarifies the formal arguments. * Add a RouterCommand + SignMachine for producing it to coins/ethereum * Ethereum which compiles * Have branch/change/forward return an option Also defines a UtxoNetwork extension trait for MAX_INPUTS. * Make external_address exclusively a test fn * Move the "account" scheduler to "smart contract" * Remove ABI artifact * Move refund/forward Plan creation into the Processor We create forward Plans in the scan path, and need to know their exact fees in the scan path. This requires adding a somewhat wonky shim_forward_plan method so we can obtain a Plan equivalent to the actual forward Plan for fee reasons, yet don't expect it to be the actual forward Plan (which may be distinct if the Plan pulls from the global state, such as with a nonce). Also properly types a Scheduler addendum such that the SC scheduler isn't cramming the nonce to use into the N::Output type. * Flesh out the Ethereum integration more * Two commits ago, into the **Scheduler, not Processor * Remove misc TODOs in SC Scheduler * Add constructor to RouterCommandMachine * RouterCommand read, pairing with the prior added write * Further add serialization methods * Have the Router's key included with the InInstruction This does not use the key at the time of the event. This uses the key at the end of the block for the event. Its much simpler than getting the full event streams for each, checking when they interlace. This does not read the state. Every block, this makes a request for every single key update and simply chooses the last one. This allows pruning state, only keeping the event tree. Ideally, we'd also introduce a cache to reduce the cost of the filter (small in events yielded, long in blocks searched). Since Serai doesn't have any forwarding TXs, nor Branches, nor change, all of our Plans should solely have payments out, and there's no expectation of a Plan being made under one key broken by it being received by another key. * Add read/write to InInstruction * Abstract the ABI for Call/OutInstruction in ethereum-serai * Fill out signable_transaction for Ethereum * Move ethereum-serai to alloy Resolves #331. * Use the opaque sol macro instead of generated files * Move the processor over to the now-alloy-based ethereum-serai * Use the ecrecover provided by alloy * Have the SC use nonce for rotation, not session (an independent nonce which wasn't synchronized) * Always use the latest keys for SC scheduled plans * get_eventuality_completions for Ethereum * Finish fleshing out the processor Ethereum integration as needed for serai-processor tests This doesn't not support any actual deployments, not even the ones simulated by serai-processor-docker-tests. * Add alloy-simple-request-transport to the GH workflows * cargo update * Clarify a few comments and make one check more robust * Use a string for 27.0 in .github * Remove optional from no-longer-optional dependencies in processor * Add alloy to git deny exception * Fix no longer optional specification in processor's binaries feature * Use a version of foundry from 2024 * Correct fetching Bitcoin TXs in the processor docker tests * Update rustls to resolve RUSTSEC warnings * Use the monthly nightly foundry, not the deleted daily nightly
2024-04-21 10:02:12 +00:00
if (coin == address(0)) {
if (amount != msg.value) revert InvalidAmount();
Ethereum Integration (#557) * Clean up Ethereum * Consistent contract address for deployed contracts * Flesh out Router a bit * Add a Deployer for DoS-less deployment * Implement Router-finding * Use CREATE2 helper present in ethers * Move from CREATE2 to CREATE Bit more streamlined for our use case. * Document ethereum-serai * Tidy tests a bit * Test updateSeraiKey * Use encodePacked for updateSeraiKey * Take in the block hash to read state during * Add a Sandbox contract to the Ethereum integration * Add retrieval of transfers from Ethereum * Add inInstruction function to the Router * Augment our handling of InInstructions events with a check the transfer event also exists * Have the Deployer error upon failed deployments * Add --via-ir * Make get_transaction test-only We only used it to get transactions to confirm the resolution of Eventualities. Eventualities need to be modularized. By introducing the dedicated confirm_completion function, we remove the need for a non-test get_transaction AND begin this modularization (by no longer explicitly grabbing a transaction to check with). * Modularize Eventuality Almost fully-deprecates the Transaction trait for Completion. Replaces Transaction ID with Claim. * Modularize the Scheduler behind a trait * Add an extremely basic account Scheduler * Add nonce uses, key rotation to the account scheduler * Only report the account Scheduler empty after transferring keys Also ban payments to the branch/change/forward addresses. * Make fns reliant on state test-only * Start of an Ethereum integration for the processor * Add a session to the Router to prevent updateSeraiKey replaying This would only happen if an old key was rotated to again, which would require n-of-n collusion (already ridiculous and a valid fault attributable event). It just clarifies the formal arguments. * Add a RouterCommand + SignMachine for producing it to coins/ethereum * Ethereum which compiles * Have branch/change/forward return an option Also defines a UtxoNetwork extension trait for MAX_INPUTS. * Make external_address exclusively a test fn * Move the "account" scheduler to "smart contract" * Remove ABI artifact * Move refund/forward Plan creation into the Processor We create forward Plans in the scan path, and need to know their exact fees in the scan path. This requires adding a somewhat wonky shim_forward_plan method so we can obtain a Plan equivalent to the actual forward Plan for fee reasons, yet don't expect it to be the actual forward Plan (which may be distinct if the Plan pulls from the global state, such as with a nonce). Also properly types a Scheduler addendum such that the SC scheduler isn't cramming the nonce to use into the N::Output type. * Flesh out the Ethereum integration more * Two commits ago, into the **Scheduler, not Processor * Remove misc TODOs in SC Scheduler * Add constructor to RouterCommandMachine * RouterCommand read, pairing with the prior added write * Further add serialization methods * Have the Router's key included with the InInstruction This does not use the key at the time of the event. This uses the key at the end of the block for the event. Its much simpler than getting the full event streams for each, checking when they interlace. This does not read the state. Every block, this makes a request for every single key update and simply chooses the last one. This allows pruning state, only keeping the event tree. Ideally, we'd also introduce a cache to reduce the cost of the filter (small in events yielded, long in blocks searched). Since Serai doesn't have any forwarding TXs, nor Branches, nor change, all of our Plans should solely have payments out, and there's no expectation of a Plan being made under one key broken by it being received by another key. * Add read/write to InInstruction * Abstract the ABI for Call/OutInstruction in ethereum-serai * Fill out signable_transaction for Ethereum * Move ethereum-serai to alloy Resolves #331. * Use the opaque sol macro instead of generated files * Move the processor over to the now-alloy-based ethereum-serai * Use the ecrecover provided by alloy * Have the SC use nonce for rotation, not session (an independent nonce which wasn't synchronized) * Always use the latest keys for SC scheduled plans * get_eventuality_completions for Ethereum * Finish fleshing out the processor Ethereum integration as needed for serai-processor tests This doesn't not support any actual deployments, not even the ones simulated by serai-processor-docker-tests. * Add alloy-simple-request-transport to the GH workflows * cargo update * Clarify a few comments and make one check more robust * Use a string for 27.0 in .github * Remove optional from no-longer-optional dependencies in processor * Add alloy to git deny exception * Fix no longer optional specification in processor's binaries feature * Use a version of foundry from 2024 * Correct fetching Bitcoin TXs in the processor docker tests * Update rustls to resolve RUSTSEC warnings * Use the monthly nightly foundry, not the deleted daily nightly
2024-04-21 10:02:12 +00:00
} else {
2024-09-17 01:34:59 +00:00
(bool success, bytes memory res) = address(coin).call(
abi.encodeWithSelector(IERC20.transferFrom.selector, msg.sender, address(this), amount)
);
Ethereum Integration (#557) * Clean up Ethereum * Consistent contract address for deployed contracts * Flesh out Router a bit * Add a Deployer for DoS-less deployment * Implement Router-finding * Use CREATE2 helper present in ethers * Move from CREATE2 to CREATE Bit more streamlined for our use case. * Document ethereum-serai * Tidy tests a bit * Test updateSeraiKey * Use encodePacked for updateSeraiKey * Take in the block hash to read state during * Add a Sandbox contract to the Ethereum integration * Add retrieval of transfers from Ethereum * Add inInstruction function to the Router * Augment our handling of InInstructions events with a check the transfer event also exists * Have the Deployer error upon failed deployments * Add --via-ir * Make get_transaction test-only We only used it to get transactions to confirm the resolution of Eventualities. Eventualities need to be modularized. By introducing the dedicated confirm_completion function, we remove the need for a non-test get_transaction AND begin this modularization (by no longer explicitly grabbing a transaction to check with). * Modularize Eventuality Almost fully-deprecates the Transaction trait for Completion. Replaces Transaction ID with Claim. * Modularize the Scheduler behind a trait * Add an extremely basic account Scheduler * Add nonce uses, key rotation to the account scheduler * Only report the account Scheduler empty after transferring keys Also ban payments to the branch/change/forward addresses. * Make fns reliant on state test-only * Start of an Ethereum integration for the processor * Add a session to the Router to prevent updateSeraiKey replaying This would only happen if an old key was rotated to again, which would require n-of-n collusion (already ridiculous and a valid fault attributable event). It just clarifies the formal arguments. * Add a RouterCommand + SignMachine for producing it to coins/ethereum * Ethereum which compiles * Have branch/change/forward return an option Also defines a UtxoNetwork extension trait for MAX_INPUTS. * Make external_address exclusively a test fn * Move the "account" scheduler to "smart contract" * Remove ABI artifact * Move refund/forward Plan creation into the Processor We create forward Plans in the scan path, and need to know their exact fees in the scan path. This requires adding a somewhat wonky shim_forward_plan method so we can obtain a Plan equivalent to the actual forward Plan for fee reasons, yet don't expect it to be the actual forward Plan (which may be distinct if the Plan pulls from the global state, such as with a nonce). Also properly types a Scheduler addendum such that the SC scheduler isn't cramming the nonce to use into the N::Output type. * Flesh out the Ethereum integration more * Two commits ago, into the **Scheduler, not Processor * Remove misc TODOs in SC Scheduler * Add constructor to RouterCommandMachine * RouterCommand read, pairing with the prior added write * Further add serialization methods * Have the Router's key included with the InInstruction This does not use the key at the time of the event. This uses the key at the end of the block for the event. Its much simpler than getting the full event streams for each, checking when they interlace. This does not read the state. Every block, this makes a request for every single key update and simply chooses the last one. This allows pruning state, only keeping the event tree. Ideally, we'd also introduce a cache to reduce the cost of the filter (small in events yielded, long in blocks searched). Since Serai doesn't have any forwarding TXs, nor Branches, nor change, all of our Plans should solely have payments out, and there's no expectation of a Plan being made under one key broken by it being received by another key. * Add read/write to InInstruction * Abstract the ABI for Call/OutInstruction in ethereum-serai * Fill out signable_transaction for Ethereum * Move ethereum-serai to alloy Resolves #331. * Use the opaque sol macro instead of generated files * Move the processor over to the now-alloy-based ethereum-serai * Use the ecrecover provided by alloy * Have the SC use nonce for rotation, not session (an independent nonce which wasn't synchronized) * Always use the latest keys for SC scheduled plans * get_eventuality_completions for Ethereum * Finish fleshing out the processor Ethereum integration as needed for serai-processor tests This doesn't not support any actual deployments, not even the ones simulated by serai-processor-docker-tests. * Add alloy-simple-request-transport to the GH workflows * cargo update * Clarify a few comments and make one check more robust * Use a string for 27.0 in .github * Remove optional from no-longer-optional dependencies in processor * Add alloy to git deny exception * Fix no longer optional specification in processor's binaries feature * Use a version of foundry from 2024 * Correct fetching Bitcoin TXs in the processor docker tests * Update rustls to resolve RUSTSEC warnings * Use the monthly nightly foundry, not the deleted daily nightly
2024-04-21 10:02:12 +00:00
// Require there was nothing returned, which is done by some non-standard tokens, or that the
// ERC20 contract did in fact return true
bool nonStandardResOrTrue = (res.length == 0) || abi.decode(res, (bool));
if (!(success && nonStandardResOrTrue)) revert FailedTransfer();
Ethereum Integration (#557) * Clean up Ethereum * Consistent contract address for deployed contracts * Flesh out Router a bit * Add a Deployer for DoS-less deployment * Implement Router-finding * Use CREATE2 helper present in ethers * Move from CREATE2 to CREATE Bit more streamlined for our use case. * Document ethereum-serai * Tidy tests a bit * Test updateSeraiKey * Use encodePacked for updateSeraiKey * Take in the block hash to read state during * Add a Sandbox contract to the Ethereum integration * Add retrieval of transfers from Ethereum * Add inInstruction function to the Router * Augment our handling of InInstructions events with a check the transfer event also exists * Have the Deployer error upon failed deployments * Add --via-ir * Make get_transaction test-only We only used it to get transactions to confirm the resolution of Eventualities. Eventualities need to be modularized. By introducing the dedicated confirm_completion function, we remove the need for a non-test get_transaction AND begin this modularization (by no longer explicitly grabbing a transaction to check with). * Modularize Eventuality Almost fully-deprecates the Transaction trait for Completion. Replaces Transaction ID with Claim. * Modularize the Scheduler behind a trait * Add an extremely basic account Scheduler * Add nonce uses, key rotation to the account scheduler * Only report the account Scheduler empty after transferring keys Also ban payments to the branch/change/forward addresses. * Make fns reliant on state test-only * Start of an Ethereum integration for the processor * Add a session to the Router to prevent updateSeraiKey replaying This would only happen if an old key was rotated to again, which would require n-of-n collusion (already ridiculous and a valid fault attributable event). It just clarifies the formal arguments. * Add a RouterCommand + SignMachine for producing it to coins/ethereum * Ethereum which compiles * Have branch/change/forward return an option Also defines a UtxoNetwork extension trait for MAX_INPUTS. * Make external_address exclusively a test fn * Move the "account" scheduler to "smart contract" * Remove ABI artifact * Move refund/forward Plan creation into the Processor We create forward Plans in the scan path, and need to know their exact fees in the scan path. This requires adding a somewhat wonky shim_forward_plan method so we can obtain a Plan equivalent to the actual forward Plan for fee reasons, yet don't expect it to be the actual forward Plan (which may be distinct if the Plan pulls from the global state, such as with a nonce). Also properly types a Scheduler addendum such that the SC scheduler isn't cramming the nonce to use into the N::Output type. * Flesh out the Ethereum integration more * Two commits ago, into the **Scheduler, not Processor * Remove misc TODOs in SC Scheduler * Add constructor to RouterCommandMachine * RouterCommand read, pairing with the prior added write * Further add serialization methods * Have the Router's key included with the InInstruction This does not use the key at the time of the event. This uses the key at the end of the block for the event. Its much simpler than getting the full event streams for each, checking when they interlace. This does not read the state. Every block, this makes a request for every single key update and simply chooses the last one. This allows pruning state, only keeping the event tree. Ideally, we'd also introduce a cache to reduce the cost of the filter (small in events yielded, long in blocks searched). Since Serai doesn't have any forwarding TXs, nor Branches, nor change, all of our Plans should solely have payments out, and there's no expectation of a Plan being made under one key broken by it being received by another key. * Add read/write to InInstruction * Abstract the ABI for Call/OutInstruction in ethereum-serai * Fill out signable_transaction for Ethereum * Move ethereum-serai to alloy Resolves #331. * Use the opaque sol macro instead of generated files * Move the processor over to the now-alloy-based ethereum-serai * Use the ecrecover provided by alloy * Have the SC use nonce for rotation, not session (an independent nonce which wasn't synchronized) * Always use the latest keys for SC scheduled plans * get_eventuality_completions for Ethereum * Finish fleshing out the processor Ethereum integration as needed for serai-processor tests This doesn't not support any actual deployments, not even the ones simulated by serai-processor-docker-tests. * Add alloy-simple-request-transport to the GH workflows * cargo update * Clarify a few comments and make one check more robust * Use a string for 27.0 in .github * Remove optional from no-longer-optional dependencies in processor * Add alloy to git deny exception * Fix no longer optional specification in processor's binaries feature * Use a version of foundry from 2024 * Correct fetching Bitcoin TXs in the processor docker tests * Update rustls to resolve RUSTSEC warnings * Use the monthly nightly foundry, not the deleted daily nightly
2024-04-21 10:02:12 +00:00
}
/*
Due to fee-on-transfer tokens, emitting the amount directly is frowned upon. The amount
instructed to be transferred may not actually be the amount transferred.
If we add nonReentrant to every single function which can effect the balance, we can check the
amount exactly matches. This prevents transfers of less value than expected occurring, at
least, not without an additional transfer to top up the difference (which isn't routed through
this contract and accordingly isn't trying to artificially create events from this contract).
If we don't add nonReentrant, a transfer can be started, and then a new transfer for the
difference can follow it up (again and again until a rounding error is reached). This contract
would believe all transfers were done in full, despite each only being done in part (except
for the last one).
Given fee-on-transfer tokens aren't intended to be supported, the only token actively planned
to be supported is Dai and it doesn't have any fee-on-transfer logic, and how fee-on-transfer
tokens aren't even able to be supported at this time by the larger Serai network, we simply
classify this entire class of tokens as non-standard implementations which induce undefined
behavior.
It is the Serai network's role not to add support for any non-standard implementations.
Ethereum Integration (#557) * Clean up Ethereum * Consistent contract address for deployed contracts * Flesh out Router a bit * Add a Deployer for DoS-less deployment * Implement Router-finding * Use CREATE2 helper present in ethers * Move from CREATE2 to CREATE Bit more streamlined for our use case. * Document ethereum-serai * Tidy tests a bit * Test updateSeraiKey * Use encodePacked for updateSeraiKey * Take in the block hash to read state during * Add a Sandbox contract to the Ethereum integration * Add retrieval of transfers from Ethereum * Add inInstruction function to the Router * Augment our handling of InInstructions events with a check the transfer event also exists * Have the Deployer error upon failed deployments * Add --via-ir * Make get_transaction test-only We only used it to get transactions to confirm the resolution of Eventualities. Eventualities need to be modularized. By introducing the dedicated confirm_completion function, we remove the need for a non-test get_transaction AND begin this modularization (by no longer explicitly grabbing a transaction to check with). * Modularize Eventuality Almost fully-deprecates the Transaction trait for Completion. Replaces Transaction ID with Claim. * Modularize the Scheduler behind a trait * Add an extremely basic account Scheduler * Add nonce uses, key rotation to the account scheduler * Only report the account Scheduler empty after transferring keys Also ban payments to the branch/change/forward addresses. * Make fns reliant on state test-only * Start of an Ethereum integration for the processor * Add a session to the Router to prevent updateSeraiKey replaying This would only happen if an old key was rotated to again, which would require n-of-n collusion (already ridiculous and a valid fault attributable event). It just clarifies the formal arguments. * Add a RouterCommand + SignMachine for producing it to coins/ethereum * Ethereum which compiles * Have branch/change/forward return an option Also defines a UtxoNetwork extension trait for MAX_INPUTS. * Make external_address exclusively a test fn * Move the "account" scheduler to "smart contract" * Remove ABI artifact * Move refund/forward Plan creation into the Processor We create forward Plans in the scan path, and need to know their exact fees in the scan path. This requires adding a somewhat wonky shim_forward_plan method so we can obtain a Plan equivalent to the actual forward Plan for fee reasons, yet don't expect it to be the actual forward Plan (which may be distinct if the Plan pulls from the global state, such as with a nonce). Also properly types a Scheduler addendum such that the SC scheduler isn't cramming the nonce to use into the N::Output type. * Flesh out the Ethereum integration more * Two commits ago, into the **Scheduler, not Processor * Remove misc TODOs in SC Scheduler * Add constructor to RouterCommandMachine * RouterCommand read, pairing with the prior added write * Further add serialization methods * Have the Router's key included with the InInstruction This does not use the key at the time of the event. This uses the key at the end of the block for the event. Its much simpler than getting the full event streams for each, checking when they interlace. This does not read the state. Every block, this makes a request for every single key update and simply chooses the last one. This allows pruning state, only keeping the event tree. Ideally, we'd also introduce a cache to reduce the cost of the filter (small in events yielded, long in blocks searched). Since Serai doesn't have any forwarding TXs, nor Branches, nor change, all of our Plans should solely have payments out, and there's no expectation of a Plan being made under one key broken by it being received by another key. * Add read/write to InInstruction * Abstract the ABI for Call/OutInstruction in ethereum-serai * Fill out signable_transaction for Ethereum * Move ethereum-serai to alloy Resolves #331. * Use the opaque sol macro instead of generated files * Move the processor over to the now-alloy-based ethereum-serai * Use the ecrecover provided by alloy * Have the SC use nonce for rotation, not session (an independent nonce which wasn't synchronized) * Always use the latest keys for SC scheduled plans * get_eventuality_completions for Ethereum * Finish fleshing out the processor Ethereum integration as needed for serai-processor tests This doesn't not support any actual deployments, not even the ones simulated by serai-processor-docker-tests. * Add alloy-simple-request-transport to the GH workflows * cargo update * Clarify a few comments and make one check more robust * Use a string for 27.0 in .github * Remove optional from no-longer-optional dependencies in processor * Add alloy to git deny exception * Fix no longer optional specification in processor's binaries feature * Use a version of foundry from 2024 * Correct fetching Bitcoin TXs in the processor docker tests * Update rustls to resolve RUSTSEC warnings * Use the monthly nightly foundry, not the deleted daily nightly
2024-04-21 10:02:12 +00:00
*/
emit InInstruction(msg.sender, coin, amount, instruction);
}
/*
We on purposely do not check if these calls succeed. A call either succeeded, and there's no
problem, or the call failed due to:
A) An insolvency
B) A malicious receiver
C) A non-standard token
A is an invariant, B should be dropped, C is something out of the control of this contract.
It is again the Serai's network role to not add support for any non-standard tokens,
*/
// Perform an ERC20 transfer out
function _erc20TransferOut(address to, address coin, uint256 value) private {
coin.call{ gas: 100_000 }(abi.encodeWithSelector(IERC20.transfer.selector, msg.sender, value));
}
// Perform an ETH/ERC20 transfer out
function _transferOut(address to, address coin, uint256 value) private {
if (coin == address(0)) {
// Enough gas to service the transfer and a minimal amount of logic
to.call{ value: value, gas: 5_000 }("");
} else {
_erc20TransferOut(to, coin, value);
Ethereum Integration (#557) * Clean up Ethereum * Consistent contract address for deployed contracts * Flesh out Router a bit * Add a Deployer for DoS-less deployment * Implement Router-finding * Use CREATE2 helper present in ethers * Move from CREATE2 to CREATE Bit more streamlined for our use case. * Document ethereum-serai * Tidy tests a bit * Test updateSeraiKey * Use encodePacked for updateSeraiKey * Take in the block hash to read state during * Add a Sandbox contract to the Ethereum integration * Add retrieval of transfers from Ethereum * Add inInstruction function to the Router * Augment our handling of InInstructions events with a check the transfer event also exists * Have the Deployer error upon failed deployments * Add --via-ir * Make get_transaction test-only We only used it to get transactions to confirm the resolution of Eventualities. Eventualities need to be modularized. By introducing the dedicated confirm_completion function, we remove the need for a non-test get_transaction AND begin this modularization (by no longer explicitly grabbing a transaction to check with). * Modularize Eventuality Almost fully-deprecates the Transaction trait for Completion. Replaces Transaction ID with Claim. * Modularize the Scheduler behind a trait * Add an extremely basic account Scheduler * Add nonce uses, key rotation to the account scheduler * Only report the account Scheduler empty after transferring keys Also ban payments to the branch/change/forward addresses. * Make fns reliant on state test-only * Start of an Ethereum integration for the processor * Add a session to the Router to prevent updateSeraiKey replaying This would only happen if an old key was rotated to again, which would require n-of-n collusion (already ridiculous and a valid fault attributable event). It just clarifies the formal arguments. * Add a RouterCommand + SignMachine for producing it to coins/ethereum * Ethereum which compiles * Have branch/change/forward return an option Also defines a UtxoNetwork extension trait for MAX_INPUTS. * Make external_address exclusively a test fn * Move the "account" scheduler to "smart contract" * Remove ABI artifact * Move refund/forward Plan creation into the Processor We create forward Plans in the scan path, and need to know their exact fees in the scan path. This requires adding a somewhat wonky shim_forward_plan method so we can obtain a Plan equivalent to the actual forward Plan for fee reasons, yet don't expect it to be the actual forward Plan (which may be distinct if the Plan pulls from the global state, such as with a nonce). Also properly types a Scheduler addendum such that the SC scheduler isn't cramming the nonce to use into the N::Output type. * Flesh out the Ethereum integration more * Two commits ago, into the **Scheduler, not Processor * Remove misc TODOs in SC Scheduler * Add constructor to RouterCommandMachine * RouterCommand read, pairing with the prior added write * Further add serialization methods * Have the Router's key included with the InInstruction This does not use the key at the time of the event. This uses the key at the end of the block for the event. Its much simpler than getting the full event streams for each, checking when they interlace. This does not read the state. Every block, this makes a request for every single key update and simply chooses the last one. This allows pruning state, only keeping the event tree. Ideally, we'd also introduce a cache to reduce the cost of the filter (small in events yielded, long in blocks searched). Since Serai doesn't have any forwarding TXs, nor Branches, nor change, all of our Plans should solely have payments out, and there's no expectation of a Plan being made under one key broken by it being received by another key. * Add read/write to InInstruction * Abstract the ABI for Call/OutInstruction in ethereum-serai * Fill out signable_transaction for Ethereum * Move ethereum-serai to alloy Resolves #331. * Use the opaque sol macro instead of generated files * Move the processor over to the now-alloy-based ethereum-serai * Use the ecrecover provided by alloy * Have the SC use nonce for rotation, not session (an independent nonce which wasn't synchronized) * Always use the latest keys for SC scheduled plans * get_eventuality_completions for Ethereum * Finish fleshing out the processor Ethereum integration as needed for serai-processor tests This doesn't not support any actual deployments, not even the ones simulated by serai-processor-docker-tests. * Add alloy-simple-request-transport to the GH workflows * cargo update * Clarify a few comments and make one check more robust * Use a string for 27.0 in .github * Remove optional from no-longer-optional dependencies in processor * Add alloy to git deny exception * Fix no longer optional specification in processor's binaries feature * Use a version of foundry from 2024 * Correct fetching Bitcoin TXs in the processor docker tests * Update rustls to resolve RUSTSEC warnings * Use the monthly nightly foundry, not the deleted daily nightly
2024-04-21 10:02:12 +00:00
}
}
/*
Serai supports arbitrary calls out via deploying smart contracts (with user-specified code),
letting them execute whatever calls they're coded for. Since we can't meter CREATE, we call
CREATE from this function which we call not internally, but with CALL (which we can meter).
*/
function arbitaryCallOut(bytes memory code) external payable {
// Because we're creating a contract, increment our nonce
_smartContractNonce += 1;
uint256 msg_value = msg.value;
address contractAddress;
assembly {
contractAddress := create(msg_value, add(code, 0x20), mload(code))
}
}
Ethereum Integration (#557) * Clean up Ethereum * Consistent contract address for deployed contracts * Flesh out Router a bit * Add a Deployer for DoS-less deployment * Implement Router-finding * Use CREATE2 helper present in ethers * Move from CREATE2 to CREATE Bit more streamlined for our use case. * Document ethereum-serai * Tidy tests a bit * Test updateSeraiKey * Use encodePacked for updateSeraiKey * Take in the block hash to read state during * Add a Sandbox contract to the Ethereum integration * Add retrieval of transfers from Ethereum * Add inInstruction function to the Router * Augment our handling of InInstructions events with a check the transfer event also exists * Have the Deployer error upon failed deployments * Add --via-ir * Make get_transaction test-only We only used it to get transactions to confirm the resolution of Eventualities. Eventualities need to be modularized. By introducing the dedicated confirm_completion function, we remove the need for a non-test get_transaction AND begin this modularization (by no longer explicitly grabbing a transaction to check with). * Modularize Eventuality Almost fully-deprecates the Transaction trait for Completion. Replaces Transaction ID with Claim. * Modularize the Scheduler behind a trait * Add an extremely basic account Scheduler * Add nonce uses, key rotation to the account scheduler * Only report the account Scheduler empty after transferring keys Also ban payments to the branch/change/forward addresses. * Make fns reliant on state test-only * Start of an Ethereum integration for the processor * Add a session to the Router to prevent updateSeraiKey replaying This would only happen if an old key was rotated to again, which would require n-of-n collusion (already ridiculous and a valid fault attributable event). It just clarifies the formal arguments. * Add a RouterCommand + SignMachine for producing it to coins/ethereum * Ethereum which compiles * Have branch/change/forward return an option Also defines a UtxoNetwork extension trait for MAX_INPUTS. * Make external_address exclusively a test fn * Move the "account" scheduler to "smart contract" * Remove ABI artifact * Move refund/forward Plan creation into the Processor We create forward Plans in the scan path, and need to know their exact fees in the scan path. This requires adding a somewhat wonky shim_forward_plan method so we can obtain a Plan equivalent to the actual forward Plan for fee reasons, yet don't expect it to be the actual forward Plan (which may be distinct if the Plan pulls from the global state, such as with a nonce). Also properly types a Scheduler addendum such that the SC scheduler isn't cramming the nonce to use into the N::Output type. * Flesh out the Ethereum integration more * Two commits ago, into the **Scheduler, not Processor * Remove misc TODOs in SC Scheduler * Add constructor to RouterCommandMachine * RouterCommand read, pairing with the prior added write * Further add serialization methods * Have the Router's key included with the InInstruction This does not use the key at the time of the event. This uses the key at the end of the block for the event. Its much simpler than getting the full event streams for each, checking when they interlace. This does not read the state. Every block, this makes a request for every single key update and simply chooses the last one. This allows pruning state, only keeping the event tree. Ideally, we'd also introduce a cache to reduce the cost of the filter (small in events yielded, long in blocks searched). Since Serai doesn't have any forwarding TXs, nor Branches, nor change, all of our Plans should solely have payments out, and there's no expectation of a Plan being made under one key broken by it being received by another key. * Add read/write to InInstruction * Abstract the ABI for Call/OutInstruction in ethereum-serai * Fill out signable_transaction for Ethereum * Move ethereum-serai to alloy Resolves #331. * Use the opaque sol macro instead of generated files * Move the processor over to the now-alloy-based ethereum-serai * Use the ecrecover provided by alloy * Have the SC use nonce for rotation, not session (an independent nonce which wasn't synchronized) * Always use the latest keys for SC scheduled plans * get_eventuality_completions for Ethereum * Finish fleshing out the processor Ethereum integration as needed for serai-processor tests This doesn't not support any actual deployments, not even the ones simulated by serai-processor-docker-tests. * Add alloy-simple-request-transport to the GH workflows * cargo update * Clarify a few comments and make one check more robust * Use a string for 27.0 in .github * Remove optional from no-longer-optional dependencies in processor * Add alloy to git deny exception * Fix no longer optional specification in processor's binaries feature * Use a version of foundry from 2024 * Correct fetching Bitcoin TXs in the processor docker tests * Update rustls to resolve RUSTSEC warnings * Use the monthly nightly foundry, not the deleted daily nightly
2024-04-21 10:02:12 +00:00
// Execute a list of transactions if they were signed by the current key with the current nonce
function execute(
address coin,
uint256 fee,
OutInstruction[] calldata transactions,
Signature calldata signature
) external {
// Verify the signature
2024-09-17 05:04:08 +00:00
// We hash the message here as we need the message's hash for the Executed event
// Since we're already going to hash it, hashing it prior to verifying the signature reduces the
// amount of words hashed by its challenge function (reducing our gas costs)
bytes32 message =
keccak256(abi.encode("execute", block.chainid, _nonce, coin, fee, transactions));
if (!Schnorr.verify(_seraiKey, message, signature.c, signature.s)) {
Ethereum Integration (#557) * Clean up Ethereum * Consistent contract address for deployed contracts * Flesh out Router a bit * Add a Deployer for DoS-less deployment * Implement Router-finding * Use CREATE2 helper present in ethers * Move from CREATE2 to CREATE Bit more streamlined for our use case. * Document ethereum-serai * Tidy tests a bit * Test updateSeraiKey * Use encodePacked for updateSeraiKey * Take in the block hash to read state during * Add a Sandbox contract to the Ethereum integration * Add retrieval of transfers from Ethereum * Add inInstruction function to the Router * Augment our handling of InInstructions events with a check the transfer event also exists * Have the Deployer error upon failed deployments * Add --via-ir * Make get_transaction test-only We only used it to get transactions to confirm the resolution of Eventualities. Eventualities need to be modularized. By introducing the dedicated confirm_completion function, we remove the need for a non-test get_transaction AND begin this modularization (by no longer explicitly grabbing a transaction to check with). * Modularize Eventuality Almost fully-deprecates the Transaction trait for Completion. Replaces Transaction ID with Claim. * Modularize the Scheduler behind a trait * Add an extremely basic account Scheduler * Add nonce uses, key rotation to the account scheduler * Only report the account Scheduler empty after transferring keys Also ban payments to the branch/change/forward addresses. * Make fns reliant on state test-only * Start of an Ethereum integration for the processor * Add a session to the Router to prevent updateSeraiKey replaying This would only happen if an old key was rotated to again, which would require n-of-n collusion (already ridiculous and a valid fault attributable event). It just clarifies the formal arguments. * Add a RouterCommand + SignMachine for producing it to coins/ethereum * Ethereum which compiles * Have branch/change/forward return an option Also defines a UtxoNetwork extension trait for MAX_INPUTS. * Make external_address exclusively a test fn * Move the "account" scheduler to "smart contract" * Remove ABI artifact * Move refund/forward Plan creation into the Processor We create forward Plans in the scan path, and need to know their exact fees in the scan path. This requires adding a somewhat wonky shim_forward_plan method so we can obtain a Plan equivalent to the actual forward Plan for fee reasons, yet don't expect it to be the actual forward Plan (which may be distinct if the Plan pulls from the global state, such as with a nonce). Also properly types a Scheduler addendum such that the SC scheduler isn't cramming the nonce to use into the N::Output type. * Flesh out the Ethereum integration more * Two commits ago, into the **Scheduler, not Processor * Remove misc TODOs in SC Scheduler * Add constructor to RouterCommandMachine * RouterCommand read, pairing with the prior added write * Further add serialization methods * Have the Router's key included with the InInstruction This does not use the key at the time of the event. This uses the key at the end of the block for the event. Its much simpler than getting the full event streams for each, checking when they interlace. This does not read the state. Every block, this makes a request for every single key update and simply chooses the last one. This allows pruning state, only keeping the event tree. Ideally, we'd also introduce a cache to reduce the cost of the filter (small in events yielded, long in blocks searched). Since Serai doesn't have any forwarding TXs, nor Branches, nor change, all of our Plans should solely have payments out, and there's no expectation of a Plan being made under one key broken by it being received by another key. * Add read/write to InInstruction * Abstract the ABI for Call/OutInstruction in ethereum-serai * Fill out signable_transaction for Ethereum * Move ethereum-serai to alloy Resolves #331. * Use the opaque sol macro instead of generated files * Move the processor over to the now-alloy-based ethereum-serai * Use the ecrecover provided by alloy * Have the SC use nonce for rotation, not session (an independent nonce which wasn't synchronized) * Always use the latest keys for SC scheduled plans * get_eventuality_completions for Ethereum * Finish fleshing out the processor Ethereum integration as needed for serai-processor tests This doesn't not support any actual deployments, not even the ones simulated by serai-processor-docker-tests. * Add alloy-simple-request-transport to the GH workflows * cargo update * Clarify a few comments and make one check more robust * Use a string for 27.0 in .github * Remove optional from no-longer-optional dependencies in processor * Add alloy to git deny exception * Fix no longer optional specification in processor's binaries feature * Use a version of foundry from 2024 * Correct fetching Bitcoin TXs in the processor docker tests * Update rustls to resolve RUSTSEC warnings * Use the monthly nightly foundry, not the deleted daily nightly
2024-04-21 10:02:12 +00:00
revert InvalidSignature();
}
// Since the signature was verified, perform execution
2024-09-17 05:04:08 +00:00
emit Executed(_nonce, message);
// While this is sufficient to prevent replays, it's still technically possible for instructions
// from later batches to be executed before these instructions upon re-entrancy
_nonce++;
Ethereum Integration (#557) * Clean up Ethereum * Consistent contract address for deployed contracts * Flesh out Router a bit * Add a Deployer for DoS-less deployment * Implement Router-finding * Use CREATE2 helper present in ethers * Move from CREATE2 to CREATE Bit more streamlined for our use case. * Document ethereum-serai * Tidy tests a bit * Test updateSeraiKey * Use encodePacked for updateSeraiKey * Take in the block hash to read state during * Add a Sandbox contract to the Ethereum integration * Add retrieval of transfers from Ethereum * Add inInstruction function to the Router * Augment our handling of InInstructions events with a check the transfer event also exists * Have the Deployer error upon failed deployments * Add --via-ir * Make get_transaction test-only We only used it to get transactions to confirm the resolution of Eventualities. Eventualities need to be modularized. By introducing the dedicated confirm_completion function, we remove the need for a non-test get_transaction AND begin this modularization (by no longer explicitly grabbing a transaction to check with). * Modularize Eventuality Almost fully-deprecates the Transaction trait for Completion. Replaces Transaction ID with Claim. * Modularize the Scheduler behind a trait * Add an extremely basic account Scheduler * Add nonce uses, key rotation to the account scheduler * Only report the account Scheduler empty after transferring keys Also ban payments to the branch/change/forward addresses. * Make fns reliant on state test-only * Start of an Ethereum integration for the processor * Add a session to the Router to prevent updateSeraiKey replaying This would only happen if an old key was rotated to again, which would require n-of-n collusion (already ridiculous and a valid fault attributable event). It just clarifies the formal arguments. * Add a RouterCommand + SignMachine for producing it to coins/ethereum * Ethereum which compiles * Have branch/change/forward return an option Also defines a UtxoNetwork extension trait for MAX_INPUTS. * Make external_address exclusively a test fn * Move the "account" scheduler to "smart contract" * Remove ABI artifact * Move refund/forward Plan creation into the Processor We create forward Plans in the scan path, and need to know their exact fees in the scan path. This requires adding a somewhat wonky shim_forward_plan method so we can obtain a Plan equivalent to the actual forward Plan for fee reasons, yet don't expect it to be the actual forward Plan (which may be distinct if the Plan pulls from the global state, such as with a nonce). Also properly types a Scheduler addendum such that the SC scheduler isn't cramming the nonce to use into the N::Output type. * Flesh out the Ethereum integration more * Two commits ago, into the **Scheduler, not Processor * Remove misc TODOs in SC Scheduler * Add constructor to RouterCommandMachine * RouterCommand read, pairing with the prior added write * Further add serialization methods * Have the Router's key included with the InInstruction This does not use the key at the time of the event. This uses the key at the end of the block for the event. Its much simpler than getting the full event streams for each, checking when they interlace. This does not read the state. Every block, this makes a request for every single key update and simply chooses the last one. This allows pruning state, only keeping the event tree. Ideally, we'd also introduce a cache to reduce the cost of the filter (small in events yielded, long in blocks searched). Since Serai doesn't have any forwarding TXs, nor Branches, nor change, all of our Plans should solely have payments out, and there's no expectation of a Plan being made under one key broken by it being received by another key. * Add read/write to InInstruction * Abstract the ABI for Call/OutInstruction in ethereum-serai * Fill out signable_transaction for Ethereum * Move ethereum-serai to alloy Resolves #331. * Use the opaque sol macro instead of generated files * Move the processor over to the now-alloy-based ethereum-serai * Use the ecrecover provided by alloy * Have the SC use nonce for rotation, not session (an independent nonce which wasn't synchronized) * Always use the latest keys for SC scheduled plans * get_eventuality_completions for Ethereum * Finish fleshing out the processor Ethereum integration as needed for serai-processor tests This doesn't not support any actual deployments, not even the ones simulated by serai-processor-docker-tests. * Add alloy-simple-request-transport to the GH workflows * cargo update * Clarify a few comments and make one check more robust * Use a string for 27.0 in .github * Remove optional from no-longer-optional dependencies in processor * Add alloy to git deny exception * Fix no longer optional specification in processor's binaries feature * Use a version of foundry from 2024 * Correct fetching Bitcoin TXs in the processor docker tests * Update rustls to resolve RUSTSEC warnings * Use the monthly nightly foundry, not the deleted daily nightly
2024-04-21 10:02:12 +00:00
for (uint256 i = 0; i < transactions.length; i++) {
// If the destination is an address, we perform a direct transfer
if (transactions[i].destinationType == DestinationType.Address) {
// This may cause a panic and the contract to become stuck if the destination isn't actually
// 20 bytes. Serai is trusted to not pass a malformed destination
(AddressDestination memory destination) =
abi.decode(transactions[i].destination, (AddressDestination));
_transferOut(destination.destination, coin, transactions[i].value);
Ethereum Integration (#557) * Clean up Ethereum * Consistent contract address for deployed contracts * Flesh out Router a bit * Add a Deployer for DoS-less deployment * Implement Router-finding * Use CREATE2 helper present in ethers * Move from CREATE2 to CREATE Bit more streamlined for our use case. * Document ethereum-serai * Tidy tests a bit * Test updateSeraiKey * Use encodePacked for updateSeraiKey * Take in the block hash to read state during * Add a Sandbox contract to the Ethereum integration * Add retrieval of transfers from Ethereum * Add inInstruction function to the Router * Augment our handling of InInstructions events with a check the transfer event also exists * Have the Deployer error upon failed deployments * Add --via-ir * Make get_transaction test-only We only used it to get transactions to confirm the resolution of Eventualities. Eventualities need to be modularized. By introducing the dedicated confirm_completion function, we remove the need for a non-test get_transaction AND begin this modularization (by no longer explicitly grabbing a transaction to check with). * Modularize Eventuality Almost fully-deprecates the Transaction trait for Completion. Replaces Transaction ID with Claim. * Modularize the Scheduler behind a trait * Add an extremely basic account Scheduler * Add nonce uses, key rotation to the account scheduler * Only report the account Scheduler empty after transferring keys Also ban payments to the branch/change/forward addresses. * Make fns reliant on state test-only * Start of an Ethereum integration for the processor * Add a session to the Router to prevent updateSeraiKey replaying This would only happen if an old key was rotated to again, which would require n-of-n collusion (already ridiculous and a valid fault attributable event). It just clarifies the formal arguments. * Add a RouterCommand + SignMachine for producing it to coins/ethereum * Ethereum which compiles * Have branch/change/forward return an option Also defines a UtxoNetwork extension trait for MAX_INPUTS. * Make external_address exclusively a test fn * Move the "account" scheduler to "smart contract" * Remove ABI artifact * Move refund/forward Plan creation into the Processor We create forward Plans in the scan path, and need to know their exact fees in the scan path. This requires adding a somewhat wonky shim_forward_plan method so we can obtain a Plan equivalent to the actual forward Plan for fee reasons, yet don't expect it to be the actual forward Plan (which may be distinct if the Plan pulls from the global state, such as with a nonce). Also properly types a Scheduler addendum such that the SC scheduler isn't cramming the nonce to use into the N::Output type. * Flesh out the Ethereum integration more * Two commits ago, into the **Scheduler, not Processor * Remove misc TODOs in SC Scheduler * Add constructor to RouterCommandMachine * RouterCommand read, pairing with the prior added write * Further add serialization methods * Have the Router's key included with the InInstruction This does not use the key at the time of the event. This uses the key at the end of the block for the event. Its much simpler than getting the full event streams for each, checking when they interlace. This does not read the state. Every block, this makes a request for every single key update and simply chooses the last one. This allows pruning state, only keeping the event tree. Ideally, we'd also introduce a cache to reduce the cost of the filter (small in events yielded, long in blocks searched). Since Serai doesn't have any forwarding TXs, nor Branches, nor change, all of our Plans should solely have payments out, and there's no expectation of a Plan being made under one key broken by it being received by another key. * Add read/write to InInstruction * Abstract the ABI for Call/OutInstruction in ethereum-serai * Fill out signable_transaction for Ethereum * Move ethereum-serai to alloy Resolves #331. * Use the opaque sol macro instead of generated files * Move the processor over to the now-alloy-based ethereum-serai * Use the ecrecover provided by alloy * Have the SC use nonce for rotation, not session (an independent nonce which wasn't synchronized) * Always use the latest keys for SC scheduled plans * get_eventuality_completions for Ethereum * Finish fleshing out the processor Ethereum integration as needed for serai-processor tests This doesn't not support any actual deployments, not even the ones simulated by serai-processor-docker-tests. * Add alloy-simple-request-transport to the GH workflows * cargo update * Clarify a few comments and make one check more robust * Use a string for 27.0 in .github * Remove optional from no-longer-optional dependencies in processor * Add alloy to git deny exception * Fix no longer optional specification in processor's binaries feature * Use a version of foundry from 2024 * Correct fetching Bitcoin TXs in the processor docker tests * Update rustls to resolve RUSTSEC warnings * Use the monthly nightly foundry, not the deleted daily nightly
2024-04-21 10:02:12 +00:00
} else {
// Prepare for the transfer
uint256 eth_value = 0;
if (coin == address(0)) {
// If it's ETH, we transfer the value with the call
eth_value = transactions[i].value;
} else {
// If it's an ERC20, we calculate the hash of the will-be contract and transfer to it
// before deployment. This avoids needing to deploy, then call again, offering a few
// optimizations
address nextAddress =
address(uint160(uint256(keccak256(abi.encode(address(this), _smartContractNonce)))));
_erc20TransferOut(nextAddress, coin, transactions[i].value);
}
// Perform the deployment with the defined gas budget
(CodeDestination memory destination) =
abi.decode(transactions[i].destination, (CodeDestination));
address(this).call{ gas: destination.gas_limit, value: eth_value }(
2024-09-17 05:04:08 +00:00
abi.encodeWithSelector(Router.arbitaryCallOut.selector, destination.code)
2024-09-17 01:34:59 +00:00
);
}
}
// Transfer to the caller the fee
_transferOut(msg.sender, coin, fee);
}
function nonce() external view returns (uint256) {
return _nonce;
}
function smartContractNonce() external view returns (uint256) {
return _smartContractNonce;
}
function seraiKey() external view returns (bytes32) {
return _seraiKey;
}
}