Add a bit of logging to mock runtime #1006

…d scripts (#504) * add devnet-m2-native network with associated features to build scripts * transitively activate m2-native on runtime * add devnet-m2-native bundle build target * fix output target for m2-native bundle * pacify clippy. * runtime: Cargo.toml sort dependencies. * Cargo.toml: point patches to master. * fix devnet-m2-bundle by propagating Cargo features. * add build preset / bundle configurations for wallaby and devnet-wasm. QoL finishing touches: - The 1:N network/features mapping in runtime/build.rs wasn't cutting it any longer. I expanded this data structure to M:N. - Added docs to runtime/build.rs. Co-authored-by: vyzo <[email protected]>

* adjust Wallaby parameters. - minimum consensus power is 16GiB. - short precommit challenge delay (10 epochs). * fix: minimum consensus power for wallaby Co-authored-by: Aayush <[email protected]>

Co-authored-by: Raúl Kripalani <[email protected]>

This change makes CI trigger whenever: - a pull request is opened/updated - a commit is pushed to `master` branch The main advantage of this setup is that it enables CI runs on PRs originating from forks.

cleanup opcode definition and resolution

- removes the explicit instruction array. - removes the undefined instructions. - removes the need for ordering instructions in a specific way. - ensures that we can't talk about invalid instructions.

* bump rust-toolchain to 1.63 [needed for const fn magic] * fix new clippy warnings, most changes is just adding `Eq` to derive, only others was avoiding reborrows and allowing a 'complex' type * change .err.expect to .expect_err, and dont bind to a unit value Co-authored-by: mriise <[email protected]>

Co-authored-by: Melanie Riise <[email protected]>

…pe (#547)

Fix compile when not compiling as a "fil actor" (e.g., for testing).

This change upgrades to the v4 bundler that _does not_ depend on the FVM to map actor types to IDs. Instead, it relies on the Type enum defined in this repo. This means new actor types can be added here without having to make changes in multiple places.

Avoid blake2b hashing in wasm Co-authored-by: Jiaying Wang <[email protected]>

Co-authored-by: lyswifter <[email protected]>

* Optimize test_vm get_actor Improves extend_sector_with_deals by 40% Signed-off-by: Jakub Sztandera <[email protected]> * Keep the cache across invocations Gives another 32% Signed-off-by: Jakub Sztandera <[email protected]> Signed-off-by: Jakub Sztandera <[email protected]>

* refactor: move opcode handlers to appropriate modules. * implement BASEFEE opcode handler. * implement SELFBALANCE opcode handler. * EVM runtime: add an initial Address type that performs ID address masking. * EVM runtime: convert EVM errors to ActorErrors. * EVM runtime: commented selfdestruct. * EVM runtime: implement DIFFICULTY, NUMBER, COINBASE opcodes. * EVM runtime: add conversion from actor ID to masked address. Make CALLER, ADDRESS, ORIGIN use masked ID addresses. * remove TODO. * add comment. * fix clippy. * adjust test for ID address flag.

* add precompiles from old branc evm-precompiles * move precompiles file, add blake2 dep with blake2f * wip call fn, use precompile error type * use u256 instead of raw bytes (cleans up code) * remove Message struct as is no longer needed * attempt getting input & output memory regions * dont use custom error type, pass in just memory instead of full execution state for memory functions * write output back into smart contract remove old TODOs * add blake2f precompile * edit TODOs * rustfmt * add modxp precompile & remove gas limit/accouting for precompiles * add expairing, use custom error type, remove gas accounting, replace unsafe code * cleanup unused things and rename precompile error variant * use fork for blakef2, organize precompiles in file * move static call type to enum * add warnings and use different slicing method * rustfmt * fix errors from merge * use InvalidMemAccess again instead of OutOfGas * lock ref-fvm to specific hash to avoid Token struct changes * go back to normal EVM precompile addresssing, unlock ref-fvm version hash * add blake2 and ec ops precompile tests * fix precompile test errors issues * Update precompiles - use methods on runtime for syscalls instead of importing SDK - fix broken tests - fix some EC methods that would expect empty bytes instead of error * implement hash and recover pubkey in tests * import sepc consts * rustfmt & clippy * Update actors/evm/src/interpreter/precompiles.rs Co-authored-by: Steven Allen <[email protected]> Co-authored-by: Steven Allen <[email protected]>

We renamed hash to hash_owned to allow for a hash_into method.

With the real FvmRuntime, we expect that cloning the &Blockstore returned by Runtime#store() would simply clone the reference. This is because inside the FVM, blockstore calls proxy out to FVM syscalls; the proxy itself holds no state, and cloning it is unimportant. However, in the MockRuntime, we were using a physical MemoryBlockstore. Cloning it would return a new physical copy, thus breaking the expectation. This commit makes MockRuntime use an Rc<MemoryBlockstore>, so all clones will point to the same MemoryBlockstore, thus preserving the expectation. Co-authored-by: Raúl Kripalani <[email protected]>

This required a small refactor to allow us to mutably borrow the Runtime.

* add etk-asm dev dependency * cargo lock * basic contract assembler * magic calc contract * rustfmt * clippy * hide error details from magic calc contract constructor * clippy * add docs for asm::new_contract

* add dispatch macros for contract assembler * avoid extraneous stack copy

* refactor evm contract invocation so that we can make cross contract calls Limits the scope of the transaction to just saving the state. * rustfmt * clean up precompile calls and make scaffolding for actual implementation - drops H160 -- wtf is that even useful - adds dispatcher for actual call implementations - removes inline directive from big things * clean up error handling surrounding call; don't panic left and right * fix error handling with status codes * implement CALL * propagate nested ActorErrors from bytecode execution * more nested actor error handling * rustfmt * pass value through CALL * rustfmt * correctly handle call variant arguments the value was missing from call/callcode. also kill the useless create call kinds. * rustfmt * update comment, move common check outside of march arm * clippy * rustfmt * fix typo * ok_or_else instead of match. * Precompile asm tests (#615) * organize test files, WIP precompile ASM test * fix precompile test asm, add tests for sha precompile * use sha2-256 instead of keccak in precompiles * un-reorg evm tests * un-rename simplecoin test * save return data * fix handling of output memory regions * address space is 32 bits * names, names, names * proxy call test * fix overflow Co-authored-by: Melanie Riise <[email protected]>

* remove H160/H256 abstraction and transactions, use (safe) Address type for addresses. * rustfmt * store addresses as a 20-byte array internally * rustfmt * implement Debug for Address * clippy * improve error message in try_from conversion failure

- Fixes the non-test build by making libsecp256k1 _optional_ unless we're testing. - Makes a lot of dependencies "test only". - Removes some unnecessary dependencies.

fixes #105

Generate a code coverage report and add it as artifact as well as uploading it to CodeCov. Co-authored-by: raulk <[email protected]> Co-authored-by: Piotr Galar <[email protected]> Co-authored-by: Alex <[email protected]>

Otherwise, rust will "unify" the features between the test targets and the non-test targets, so we'll miss non-test dependency issues.

* add note about gas * barf with jumbo output addresses; can't do more than 32 bits anyway in wasm. * add TODO comment about output memory limit

* EVM runtime: implement GetBytecode and GetStorageAt getters. GetBytecode returns the CID of the bytecode block. In the future, the FVM's reachability analysis should add this CID to the reachable set of the caller. GetBytecode is publicly callable. GetStorageAt returns the value stored in the supplied EVM storage key. This method is private and only intended for calls from instrumentation and client functionality, such as the eth_getStorageAt JSON-RPC API operation. In order to make it private, it validates that the caller is the zero address. It needs to be private to preserve EVM guarantees: no contract is able to arbitrarily access storage keys from another contract. The client must form a message from the 0x0 sender address for this method to work. * it's not scala! Co-authored-by: vyzo <[email protected]>

* implement CALLACTOR/METHODNUM * signal EVM_CONTRACT_REVERTED error on reverts * handle REVERT in CALL/CALLACTOR result * update comments about REVERT semantics * simplify call error handling * fix tests * fix empty input data handling for calls * add mnemonic opcodes for CALLACTOR and METHODNUM to assembler such a hack, but no other way. * add methodnum and callactor tests * rustfmt * lift return * fix comment

Co-authored-by: vyzo <[email protected]>

* chore: run tests with some optimizations enabled Specifically, optimize bigint, hashing, and the test_vm. Preferiably we'd just optimize everything, but that makes things slow. * ci: save test cache * ci: don't build root bundle in tests We test when building the bundle itself. * ci: rev key and save codecov cache Co-authored-by: ZenGround0 <[email protected]>

This can cause issues with caching. Unfortunately, we can only specify them per-target, not per-profile.

This isn't as big of an issue as hashing, but it still matters for tests.

Resolves #567

* fix broken init test * fix broken rand import * only enable rand/std_rng when test_utils is specified

EVM contracts don't take parameters on construction.

rename evm actor's type to

* feat: use consts for addrs And expose actor IDs so we don't need to call `.id().unwrap()` to retrieve them. Co-authored-by: mriise <[email protected]>

Co-authored-by: lyswifter <[email protected]>

…sts (#654) * bn pair tests and errors * add ec_recover test, fix broken implementation * add modexp test, add more test cases for ec_recover * only check lower bound for ec_recover * Fix broken failure modes in precompiles - ec_recover - modexp - ec_add - ec_mul Change error enum to include field error types + conversions - forked bn to derive Eq on bn Errors Add helper for reading padded bytes from input. * rustfmt & clippy * Update precompiles.rs * use cow instead of custom container * remove fork of bn * remove bn for dep * allocate ahead of time when reading sets of * lazy_static secp256k1 lower bound * left pad ecrecover * hex literal and cargo lock * add padding test * space * dont panic on modexp len * few remaining nits * rustfmt * better read logic of bigints for modexp * add notes to reading bigint lengths & refactor read group for more clarity * rustfmt

The embryo actor: - Has no constructor (the system is expected to simply "set" the embryo code CID, not construct it). - Immediately aborts when called with a non-zero method. We could, alternatively, use some well-known "id" address. But it's nice to make this actual valid wasm.

Otherwise, we'll perform a hash in WASM when we first dereference (max once per call) which is just wasteful. This also fixes the integration tests to use the "correct" empty object. Previously, they were using `()` which mapped to null, not `[]`.

Previously, we weren't taking the nonce into account which prevented us from testing multiple messages from the same account.

- Adds a create4 method to the init actor. - Adds embryo actor creation to the test vm.

Co-authored-by: dignifiedquire <[email protected]>

* EVM-887: Makefile to generate Solidity compilation artifacts. * EVM-887: Use solc 0.8.15 * EVM-887: Storage test contract * EVM-887: Fix contract name. * EVM-887: Just return the sum without writing it to the counter.

* EVM-888: ABI bindings and a default instance of it. * EVM-888: TestEnv with deploy. * EVM-888: Test that a call can be invoked and the results parsed. * EVM-888: Move TestEnv out. * EVM-888: Static init for contract. * EVM-888: Capture storage stats. * EVM-888: Push at least 1, otherwise storage size is constant. * EVM-888: Export stats. * EVM-888: Add two series. * EVM-888: Fix clippy warnings. * EVM-888: Test mapping add. * EVM-888: Measure array read. * EVM-888: Change i to go from 0 instead of 1 * EVM-888: Measure mapping read. * EVM-888: Assert sums. * EVM-888: Counter increments. * EVM-888: Counter after fill. * EVM-888: Just consume the call, lose the closure. * EVM-888: Overwrite * EVM-888: Rename back to inc * EVM-888: Use the existing TrackingBlockstore * EVM-888: Undo line breaks in toml * EVM-888: Use std::env! * EVM-888: Convert to EthAddress first.

* EVM-889: Visualize measurements * EVM-889: Check in rendered charts. * EVM-889: Convenience target to run footprint measurements. * EVM-889: Separate script for plotting

…des (#713) Co-authored-by: Raúl Kripalani <[email protected]>

This updates fvm_sdk and fvm_shared to the latest released version ahead of some f4 changes (so we can atomically back out of those changes).

feat: save predictable addresses in the actor tree

* chore: update fvm encoding This switches to a bundled "strict_bytes" module instead of the "cs_serde_bytes" crate. This: - cs_serde_bytes was already a fork of serde_bytes. - We only needed a very small fraction of this crate. - We can now support fixed byte arrays. * refactor: import strict_bytes instead of serde_bytes

Co-authored-by: Steven Allen <[email protected]>

- Add a new type. - Add it to the bundle. - Add it to the test VM.

Part of #973.

* fix: encode/decode EVM parameters as CBOR The previous code was passing parameters as raw bytes, but telling the system that these bytes were valid CBOR. Unfortunately: - This is incorrect and will break once we actually start enforcing this. - This makes it _very_ difficult to test the EVM because we have quite a bit of testing code that assumes CBOR. The long-term solution is to correctly handle IPLD (see #758). But the short-term solution is to just encode/decode parameters as CBOR. * fix: only use RawBytes for CBOR "RawBytes" is for storing serialized CBOR (per the docs). Any other uses are bugs.

* feat: refactor the EVM to manage state in a consistent location - Deduplicate code between the constructor and call. - Add logic for flushing/reloading state. - Add a helper that flushes, sends, and reloads. This also fixes the whole reborrow mess. Through mostly trial and error, I found that a few opcodes were causing the problem. Fixing those opcodes made everything happy. * fix: save/restore EVM contract state before/after sending messages * fix: remove needless lifetimes Apparently, the real issue was that these lifetimes were over constrained. * chore: remove unused type * chore: add some docs * fix: use the correct nonce The EVM _used_ to start the nonce at zero, but contracts now start the nonce at one.

* fix: implement Cbor for Create2/Create params/return * test: add an EVM call test

(for recent changes)

- call - callactor - balance - selfdestruct - origin - address - get_evm_bytecode Also remove some of the "from ID" functions to reduce the chances that we'll mess this up.

* implement STATICCALL Also, consistently rename platform argument to system. * implement DELEGATECALL * rustfmt * move readonly check in flush * shortcut reload in read only mode * dont store the delegate bytecode * input parameters for delegate call are raw bytes * fix caller validation check * dedup bytecode loading code * cliptarded * WIP: callvariant tests * callvariant tests * assert actor type * fix tests * fix test contract: uncomment accidentally commented dispatch code * fix contract bugz * STATICCALL tests * fix contract bugz * delegate call tests * rustfmt * fix: handle empty bytecode on delegatecall and refactor a bit * chore: rename delegate -> with_code * fix clippy Co-authored-by: Steven Allen <[email protected]>

* fix: allow calling more non-contract actors - Allow calling accounts. - Allow calling embryos by ID. And oxidize it a bit. * fix: really fix contract calling rules 1. Don't do any extra work for dynamic calls. 2. Avoid creating new actors unless we're sending value (EVM rules). 3. Correctly identify "non-existent" actors. 4. Correctly handle "empty" returns. * test: check send/invoke logic

This is what the EVM does. They're defined to take the low 20 bytes.

This will allow lotus to call this from the fake embryo accounts. This isn't behind a feature gate because it should be secure regardless.

- byte: use the built-in byte and don't touch u128s. - sar: simplify a bit - remove dead code

This should significantly speedup smart contracts.

For a better testnet experience.

* drop m2-native from wallaby build * rustfmt

We use the ETH address derived from the user's ID address.

Users shouldn't _naturally_ try to create actors with these addresses, but it's not theoretically impossible to _make_ this happen.

Always encode to a big-endian byte array. fixes filecoin-project/ref-fvm#912

* EVM: simplify and fix evm storage 1. The EVM doesn't distinguish between "exists" and "doesn't exist". 2. There was some interesting dead code. 3. Only mark the state-tree dirty if something has changed. * nit: make vyzo happy

We only delegate methods >= 1024 to the contract. This is more than sufficient to cover any methods exported by FRC0042. part of filecoin-project/ref-fvm#1056

…ABI (#834) * on unhandled methods, call filecoin_native_method with solidity ABI * address feedback * comment * fix compile otherwise, we conflict with ActorCode::invoke_method * fix tests * fix: rename method number

* FVM specific EVM precompiles - changes randomness functions to take a raw i64 since this is userspace now - add precompiles: resolve_address, lookup_address, get_actor_code_cid, get_randomness(beacon/chain) * leave context as TODO * fixes after rebase * rustfmt * add new user randomness functions (that dont limit types to domain seperation tags) * add some precompile docs, use ID addresses instead of ETH style ID addresses, assert reserved bytes are zeroes * fmt & clippy * Update precompiles.rs * kill old expect randomness functions and route them through the new ones * rename user get_randomness_x functions to be more consistient * rustfmt

* use more EVM style params for precompiles

…P-0049) (#839)

- CALL: truncates the return value, but doesn't zero fill. - CALLDATA, EXTCODECOPY, CODECOPY: treat input as if it were followed by infinite zeros. - RETURNDATACOPY: explicitly forbids out-of-bounds reads. We might be able to change the behavior of EXTCODECOPY and CODECOPY to match RETURNDATACOPY, but we _can't_ change CALLDATA as solidity abuses this feature for zeroing memory. So we're just going to match what the EVM does because it's safer (and because we need to implement that behavior _anyways_ for CALLDATA). fixes filecoin-project/ref-fvm#1021 fixes filecoin-project/ref-fvm#1024

We ran out of space.

We can already handle _executing_ with empty bytecode, we just didn't allow construction.

Includes refactored network/message context syscalls.

* wip callactor precompile (very broken) * fill out logic a bit * rustfmt * pass in context through to precompiles * exit with error on static call to callactor, array chunks abstraction * forgot increment cursor... * actually use gas from call, assert unused bits in randomness params are zeroed * pass precompile context to all precompiles add a chunked parameter reader * improve parameter reader, check for readonly, new precompiles only read u32 for dynamicly sized bytes * update sdk * review changes, fix fn name in runtime * nit: don't comment out code

* add exit to runtime interface * propagate revert data in contract invocations * add data to ActorError * propagate data in actor errors from send * correctly handle reverts in calls * implement exit with panics in mock runtime * implement exit with panics in test vm * fix clippy * fix test vm: exit data should be in new context * use pop instead of ugly gets in ret * cosmetics * add naked revert test * rustfmt * only test naked revert in unit test nested call revert should go in integration test, really * fix callvariants contract to check reverts * really fix callvariants contract: iszero is what you want * test_vm: missing panic handler * callvariants is also testing mutation * deduplicate actor exit handler code * rustfmt * remove unnecessary replaces for exits. * fix runtime trampoline to propagate data from errors * perform contract/constructor invocation abortive returns with errors * don't synethesize exit data if it is not there * propagate error data in test_vm * rustfmt * simplify/dedup identical code

* clean up leftovers from CALLACTOR and METHODNUM refactor. * fix callactor test * rusftm Co-authored-by: mriise <[email protected]>

* add unit tests for some opcodes ADD MUL SUB DIV * fmt

* fix call_actor precompile; modify error message * revert modifying error message * rustfmt

* check stack limits at dispatch, make stack ops unchecked * use a growable vector for the stack * fix growth check * clippy wants default... * optimize ensure somewhat * remove implicit assumption about growth patterns * drop error from Bytecode::new signature, it can't really err. fix rebase artifact * refactor part I: primops * add arity/stack checks for primops * dup and swap * push * immediates * sugar * first stdfun * keccak256 as stdfun and prettier sugar * moar stdfuns * calldatacopy * emacs auto-indent friendly sugar * more functionoids * moar funcionoids * call opcodes also kill leftover callactor * primitive -> primop * codesize and codecopy * create and create2 * control flow * mark stack ops as unsafe only macro mechanically checked invocations for these * fix tests * rustfmt * shut up clippy's frivolous complaints * nit: macro magic Co-authored-by: Steven Allen <[email protected]>

* defeat instruction zoo call convention complexity it's just one, the intrinsic, which passes a machine. * fix tests * rip ControlFlow enum * more macrology * cosmetics * shut up clippy

* Update FVM * feat: correctly handle "read-only" errors. Co-authored-by: Steven Allen <[email protected]>

* fix: correctly compute addresses in CREATE And add some basic tests. fixes #866 * Apply suggestions from code review Co-authored-by: Melanie Riise <[email protected]> Co-authored-by: Melanie Riise <[email protected]>

* Export Datacap Actor methods * Export Init Actor methods * Export Market Actor methods * Export Miner Actor methods * Export Multisig Actor methods * Export Verifreg Actor methods * Address review

* Power actor: Add exported getters for raw power * FRC-XXXX is FRC-0042 * Power actor: network_raw_power: Return this_epoch_raw_byte_power * Power actor: miner_raw_power: Return whether above consensus min power * Power actor: types: serialize one-element structs transparently * Address review * Miner actor: Add exported getters for info and monies (#811) * Miner actor: Add exported getters for info and monies * Tweak comment * Miner actor: Replace GetWorker and GetControls with IsControllingAddress * Miner actor: Add exported GetAvailableBalance * Miner actor: Add exported GetVestingFunds * Miner actor: Remove exported monies getters * Miner actor: types: serialize one-element structs transparently * Address review * Address review

Co-authored-by: zenground0 <[email protected]>

* Market actor: GetDealTermExported: Return (start_epoch, duration) * Market actor: Export getter for deal total price

…#824) * Paych actor: Drop account req, use AuthenticateMessage to verify sigs * Address review * Address review

* Miner: Export ChangeWorkerAddress * Miner: Export ChangePeerID * Miner: Export WithdrawBalance * Miner: Export ChangeMultiaddrs * Miner: Export ConfirmUpdateWorkerKey * Miner: Export RepayDebt * Miner: Export ChangeOwnerAddress * Miner: Add exported getters for PeerID & multiaddrs * Miner: Refactor: Rename ConfirmUpdateWorkerKey to ConfirmChangeWorkerAddress

* Power: Export CreateMiner * Power Actor: Export MinerCount and MinerConsensusCount * Update actors/power/src/lib.rs Co-authored-by: Alex <[email protected]> Co-authored-by: Alex <[email protected]>

* Verifreg: Rename AddVerifierClientParams to AddVerifiedClientParams * Verifreg: Export AddVerifiedClient and GetClaims

We exposed this for the GASLIMIT opcode, but didn't actually need it. I'm removing it here to reduce the diff and API surface area. See filecoin-project/ref-fvm#1168 for context.

@vyzo

(@vyzo 👀)

* feat: correctly implement staticcall * fix: correctly handle staticcall in callactor * fix: clippy

* use gas limit for sends when supplied * implement send_with_gas in mock runtime and integration framework * fix read_only flag determination * fix test_vm handling of read only is send_with_gas * use U256::to_u64_saturating instead of panicky as_u64 * remove unnecessary sticky bit check * rustfmt

* wip fixes * cleanup implementation * return 0 for non existient address in extcodesize * use enum instead of result result maddness * fmt & cleanup some logic * update tests for the remaining two ext opcodes * redo logic to make things a bit more clearn * fix bugs & keep bytecode logic seperate, verify expectations before asserting for better debugging help * add another test case for ext op * use closure for early return * nit: reduce indentation * fix: formatting * remove unnecessary closure * fix: make delegatecall behave more like the EVM - address todo - use the new get_cid_type * fix some review nits Co-authored-by: Steven Allen <[email protected]>

Unfortunately, this means we need to pass the "system" to the precompiles, which means we need an unsafe dance to pass a mutable pointer through. But it's not _too_ bad.

thread gas go delegatecall

* use while insteas of loop * remove unnecessary bounde checking * fix typo * rustfmt

1. Lift the "is none" check into the loop. This is actually _slower_ for short programs, but faster over the long-run. I measured that by moving the check _above_ the step operation and determining that it was slower. 2. Remove the unsafe and inline. 1. Inlining is faster overall (at least in terms of gas usage). 2. Inlining lets rust elide the bounds checks (no difference in gas usage). 3. Don't bother checking instructions. Instead, we just dispatch and let run the "undefined instruction" function on undefined instructions. It gives the same result, but saves some time.

1. Avoid `as_usize` entirely. 2. Add some more explicit casts and overflow checking. I still hate it, but I now hate it a bit less. I'd love to get rid of the memory region abstraction, or replace it with something a little less brittle. fixes filecoin-project/ref-fvm#1185

* jmp instructions uncondintionally return the next pc No need for the option and the branches that come from it * jmp instructions can skip the noop sled saves a noop cycle, as it is guaranteed to be there.

This is legal in the EVM.

My goal here was simplify the logic around "exiting" so that I could reason about it. Specifically, so I could fix some issues with SELFDESTRUCT. We had multiple error and success paths and that made me nervous. In the process, I ended up changing a bunch of other things: 1. Exit early by setting the PC. 2. Go back to propagating the error through `step()` (makes things easier to read). 3. Fix some rust unsafety things. 4. Move _all_ of the instruction specific logic into the instruction definitions. This will let us eventually remove one of these tables. - This means that instructions manage their own PCs. 5. Aggressively inline.

This will be _properly_ fixed in filecoin-project/ref-fvm#987, but we need to return proper CBOR till then.

In the EVM: 1. Self destruct continues even if it sends funds into oblivion. 2. Self destruct will auto-create a beneficiary. This lets us punt filecoin-project/ref-fvm#736 to M2.2.

value + magic gas limit 23k implies METHOD_SEND

* fix: return data from callactor on revert as well Also, avoid cloning unnecessarily. * test: callactor return data on revert

* optimize push for u64 * somewhat nicer macro expansion

* reserve gas for return on CALL better handling of edge cases handle 0 gas limit correctly, we still have to reserve simplify * don't be sexist * fix tests; we call gas_available now on every call with gas 0. * user gas is not sacred it turns out * use gas avail for min when user specifies gas * take user's thing if it already reserves. * use min * lift common code

* wip * exit early on insufficient balance * rustfmt

* rewrite get_actor_code_cid to get_actor_type * fix review comments & write tests * fix more review its * update tests and use a helper fuction

split precompiles into smaller, more readable files

* Remove the market actor state mutation pattern (#734) Fixes #656 * Proof of concept exported API for Account actor (#797) * Export stable methods for public access (#807) * Export Datacap Actor methods * Export Init Actor methods * Export Market Actor methods * Export Miner Actor methods * Export Multisig Actor methods * Export Verifreg Actor methods * Address review * Restrict internal APIs of all actors (#809) * Exported API method for market actor escrow/locked balance (#812) * Power actor: Add exported getters for raw power (#810) * Power actor: Add exported getters for raw power * FRC-XXXX is FRC-0042 * Power actor: network_raw_power: Return this_epoch_raw_byte_power * Power actor: miner_raw_power: Return whether above consensus min power * Power actor: types: serialize one-element structs transparently * Address review * Miner actor: Add exported getters for info and monies (#811) * Miner actor: Add exported getters for info and monies * Tweak comment * Miner actor: Replace GetWorker and GetControls with IsControllingAddress * Miner actor: Add exported GetAvailableBalance * Miner actor: Add exported GetVestingFunds * Miner actor: Remove exported monies getters * Miner actor: types: serialize one-element structs transparently * Address review * Address review * Built-in market API for deal proposal metadata (#818) * Call exported authenticate method from PSD (#829) Co-authored-by: zenground0 <[email protected]> * Drop CALLER_TYPES_SIGNABLE and signable caller validation (#821) * Market actor: Minor improvements to two exported getters (#826) * Market actor: GetDealTermExported: Return (start_epoch, duration) * Market actor: Export getter for deal total price * Exported API for market deal activation state (#819) * Paych actor: Drop account req, use AuthenticateMessage to verify sigs (#824) * Paych actor: Drop account req, use AuthenticateMessage to verify sigs * Address review * Address review * Account actor: Deprecate AuthenticateMessage (#856) * Market actor: Export PublishStorageDeals (#857) * Miner: Export several more methods (#863) * Miner: Export ChangeWorkerAddress * Miner: Export ChangePeerID * Miner: Export WithdrawBalance * Miner: Export ChangeMultiaddrs * Miner: Export ConfirmUpdateWorkerKey * Miner: Export RepayDebt * Miner: Export ChangeOwnerAddress * Miner: Add exported getters for PeerID & multiaddrs * Miner: Refactor: Rename ConfirmUpdateWorkerKey to ConfirmChangeWorkerAddress * Power actor: Export methods to CreateMiner and get miner counts (#868) * Power: Export CreateMiner * Power Actor: Export MinerCount and MinerConsensusCount * Update actors/power/src/lib.rs Co-authored-by: Alex <[email protected]> Co-authored-by: Alex <[email protected]> * Verifreg: Export AddVerifiedClient and GetClaims (#873) * Verifreg: Rename AddVerifierClientParams to AddVerifiedClientParams * Verifreg: Export AddVerifiedClient and GetClaims * Datacap actor: Modify exported methods (#909) * Datacap: Export Mint and Destroy * Datacap actor: Deprecate all internal methods * Datacap actor: Rename BalanceOf to Balance * Datacap actor: Add Granularity method * fix: comments on newly exported methods (#910) Co-authored-by: RK <[email protected]> Co-authored-by: Alex <[email protected]> Co-authored-by: ZenGround0 <[email protected]> Co-authored-by: zenground0 <[email protected]>

* EVM-899: Skip write when no change (#709) * EVM-899: Don't even call the HAMT store if there was no change. * EVM-899: Measurement updates with flush fix. * EVM-890: No hashing by HAMT (#730) * EVM-890: Use the identity hasher in the EVM * EVM-890: Update the measurements. * EVM-890: Fix newline in bash script. * EVM-890: Fix charts missing series. * EVM-890: Add key wrapper and use big endian order in hash. * EVM-890: Update measurements with big endian key order. * EVM-890: Remove leftover let binding. * EVM-890: Derive serde * EVM-890: Fix jq check. * EVM-891: HAMT link extension measurements (#739) * EVM-891: Use extensions. * EVM-891: Update measurements * EVM-891: Point at the PR branch. * EVM-891: Update measurements after single char fields * EVM-890: Fix jq check. * EVM-892: Keep root small (#740) * EVM-892: Keep the first 2 levels free of data. * EVM-892: Update patch branch * EVM-892: Reduce bit width to 5 (#741) * EVM-1085: Use Kamt (#855) * EVM-1085: Use Kamt * EVM-1085: Update measurements * EVM-1085: Use the kamt branch on CI. * EVM-1085: Remove use_extensions option.

* turn gasless value transfers and 2300s into sends * fix gas in callvariants * add test for send transformation when value and no gass * add test for send transformation when no value and gas is 2300 * clippy

* wip * add sending with flags to runtime * fix a few different issues - expectations of send flags - call test - ensure readonly is being passed through * replicate sticky readonly behavior of the FVM in the test VM * add value as a parameter to call_actor precompile * Update actors/evm/src/interpreter/instructions/call.rs Co-authored-by: Steven Allen <[email protected]> * fix issues from review - static mode violations are handled by FVM now, remove internal logic around it - read_only in test vm is "context" which is/will probably be different than the _Context_ bits * restrict default impl of to just tests * oops * fmt Co-authored-by: Steven Allen <[email protected]>

1. Update the FVM's SDK to the version that treats 0 as 0 instead of infinity. 2. Correctly set the gas limit to infinity if we're demoting a call to METHOD_SEND. 3. If the user passes 0 gas in the EVM, leave it at 0 (except when demoting to METHOD_SEND).

This removes the send_read_only variant and makes the simplified `send` variant just call `send_generalized.

* Export Datacap Actor methods * Export Init Actor methods * Export Market Actor methods * Export Miner Actor methods * Export Multisig Actor methods * Export Verifreg Actor methods * Address review

* Power actor: Add exported getters for raw power * FRC-XXXX is FRC-0042 * Power actor: network_raw_power: Return this_epoch_raw_byte_power * Power actor: miner_raw_power: Return whether above consensus min power * Power actor: types: serialize one-element structs transparently * Address review * Miner actor: Add exported getters for info and monies (#811) * Miner actor: Add exported getters for info and monies * Tweak comment * Miner actor: Replace GetWorker and GetControls with IsControllingAddress * Miner actor: Add exported GetAvailableBalance * Miner actor: Add exported GetVestingFunds * Miner actor: Remove exported monies getters * Miner actor: types: serialize one-element structs transparently * Address review * Address review

Co-authored-by: zenground0 <[email protected]>

* Market actor: GetDealTermExported: Return (start_epoch, duration) * Market actor: Export getter for deal total price

…#824) * Paych actor: Drop account req, use AuthenticateMessage to verify sigs * Address review * Address review

* Miner: Export ChangeWorkerAddress * Miner: Export ChangePeerID * Miner: Export WithdrawBalance * Miner: Export ChangeMultiaddrs * Miner: Export ConfirmUpdateWorkerKey * Miner: Export RepayDebt * Miner: Export ChangeOwnerAddress * Miner: Add exported getters for PeerID & multiaddrs * Miner: Refactor: Rename ConfirmUpdateWorkerKey to ConfirmChangeWorkerAddress

* Power: Export CreateMiner * Power Actor: Export MinerCount and MinerConsensusCount * Update actors/power/src/lib.rs Co-authored-by: Alex <[email protected]> Co-authored-by: Alex <[email protected]>

* Verifreg: Rename AddVerifierClientParams to AddVerifiedClientParams * Verifreg: Export AddVerifiedClient and GetClaims

* Datacap: Export Mint and Destroy * Datacap actor: Deprecate all internal methods * Datacap actor: Rename BalanceOf to Balance * Datacap actor: Add Granularity method

This reverts commit b35e0e5.

Also, always don't abort/revert if we fail to decode the revert data. Instead, just return the data without decoding it. This shouldn't happen if we call an EVM contract, but may happen if we call another contract.

* dont revert on invalid call to exthash * EXT opcode fixes: - add GetBytecodeHash method - store bytecode hash, keccak256(bytecode) in state - extcodehash return empty hash for precompiles and keccak256([0xfe]) for native actors * update storage benchmark outputs * return empy hash for accounts, newtype for evm bytecode hash values * re-update benchmarks

* add some tests and be exaustive with match arms * fix test comment

Co-authored-by: zenground0 <[email protected]>

* wip * cache result of randomness inside of interpeter add notes about randomness values in filecoin add tests for prevrandao change evm prevrandao domain tag to 10 * review fixes * fmt

* fix: reject (f4) actor redeploy after selfdestruct We could also reject this in the FVM, but then we'd have to tell the FVM that we're re-assigning an existing ID address. The downside of this method is that it's _slightly_ more expensive (we have to make multiple syscalls instead of one). The benefit is that it makes auditing and testing this code _much_ simpler. Before, we were enforcing half the constraints in the FVM, and the other half in the init actor, which is why we had into this bug in the first place. fixes filecoin-project/ref-fvm#1175 * fixes * test it * fix compile

* Fix caller check in EVM/EAM constructors

Update context.rs

…970) re-order address and data size params of call_actor precompile

* dont special case 'system' id addresses, always check type * fix actor type precompile test * fix panic in actor type, add test for it

This is just dead code at this point. fixes filecoin-project/ref-fvm#1292

update embroy actor to placeholder according to filecoin-project/FIPs#577

) * fix: EVM: CALLs to EthAccount should be lowered to bare transfers. * catch another one.

* move native precompiles to their own addresses with prefix 0xfe * fmt and remove some stale TODOs * fix review comments * comment style consisticy

* fix evm_delegate_call's caller bug * use Eth address for delegate call params instead of ID address * fmt, add test for delegatecall Co-authored-by: samuerio <[email protected]>

* disable get-randomness precompile * only remove precompile

* Use system.send instead of system.rt.send to properly flush state before calling into another actor. This is needed since ext* opcodes can/will be called during initcode. Ext* calling self durin initcode must have _some_ state flushed in order to fetch it, system.send will flush empty bytecode hashes to state before send for us. * fmt & fix tests * add tests for ext size/hash in initcode * clippy * fix delegate_call test

update EAM to check for new reserved range for native precompiles addresses

* check that precompile address last byte/index is within range of precompile array to avoid panic * clippy

And add a bunch of tests.

* move error handling logic into call instead of from trait, remove unused CallActorError * use err instead of flattening to empty vec right away * fmt * add more logging to precompile calls * add test for empty failure * fmt

* account for null and native precompile address in eth to f4 address conversions * fmt & err message fix * be explicit about rejecting null address in EAM * add eam test for null address

Commits on Jul 22, 2022

[m2 native] user actor installation (#491 )

vyzo authored Jul 22, 2022

Configuration menu

View commit details

Copy full SHA for a87c640

Browse repository at this point

Copy the full SHA

a87c640 View commit details

Browse the repository at this point in the history

Commits on Oct 22, 2022

upgrade to [email protected].

raulk committed Oct 22, 2022

Configuration menu

View commit details

Copy full SHA for cac1981

Browse repository at this point

Copy the full SHA

cac1981 View commit details

Browse the repository at this point in the history

Commits on Nov 14, 2022

upgrade fvm dependencies. (#840 )

raulk authored Nov 14, 2022

Configuration menu

View commit details

Copy full SHA for e8d71b2

Browse repository at this point

Copy the full SHA

e8d71b2 View commit details

Browse the repository at this point in the history

Add a bit of logging to mock runtime #1006

Are you sure you want to change the base?

Add a bit of logging to mock runtime #1006

Commits on Jul 22, 2022

Commits on Jul 28, 2022

Commits on Aug 18, 2022

Commits on Aug 19, 2022

Commits on Aug 21, 2022

Commits on Aug 23, 2022

Commits on Aug 24, 2022

Commits on Aug 29, 2022

Commits on Aug 31, 2022

Commits on Sep 1, 2022

Commits on Sep 2, 2022

Commits on Sep 5, 2022

Commits on Sep 6, 2022

Commits on Sep 7, 2022

Commits on Sep 8, 2022

Commits on Sep 9, 2022

Commits on Sep 10, 2022

Commits on Sep 12, 2022

Commits on Sep 13, 2022

Commits on Sep 14, 2022

Commits on Sep 15, 2022

Commits on Sep 16, 2022

Commits on Sep 17, 2022

Commits on Sep 20, 2022

Commits on Sep 21, 2022

Commits on Sep 22, 2022

Commits on Oct 1, 2022

Commits on Oct 3, 2022

Commits on Oct 11, 2022

Commits on Oct 12, 2022

Commits on Oct 14, 2022

Commits on Oct 15, 2022

Commits on Oct 17, 2022

Commits on Oct 19, 2022

Commits on Oct 20, 2022

Commits on Oct 21, 2022

Commits on Oct 22, 2022

Commits on Oct 24, 2022

Commits on Oct 25, 2022

Commits on Oct 31, 2022

Commits on Nov 1, 2022

Commits on Nov 8, 2022

Commits on Nov 9, 2022

Commits on Nov 11, 2022

Commits on Nov 12, 2022

Commits on Nov 13, 2022

Commits on Nov 14, 2022

Commits on Nov 15, 2022

Commits on Nov 16, 2022

Commits on Nov 17, 2022

Commits on Nov 18, 2022

Commits on Nov 22, 2022

Commits on Nov 23, 2022

Commits on Nov 24, 2022

Commits on Nov 28, 2022

Commits on Nov 29, 2022

Commits on Nov 30, 2022

Commits on Dec 1, 2022

Commits on Dec 2, 2022

Commits on Dec 3, 2022

Commits on Dec 5, 2022

Commits on Dec 6, 2022

Commits on Dec 7, 2022

Commits on Dec 8, 2022

Commits on Dec 11, 2022

Commits on Dec 12, 2022

Commits on Dec 13, 2022

Commits on Dec 14, 2022

Commits on Dec 15, 2022

Commits on Dec 16, 2022

Commits on Dec 17, 2022

Commits on Dec 19, 2022

Commits on Dec 20, 2022

Commits on Dec 21, 2022

Commits on Dec 22, 2022

Commits on Jan 3, 2023

Commits on Jan 4, 2023