Implement Simple Subroutines for the EVM (EIP 2315) (openethereum#11629)

* Implement Simple Subroutines for the EVM (EIP 2315) * Finished base impl, added tests * Add EIP2315 into berlin chainspec * Update ethcore/evm/src/interpreter/shared_cache.rs Co-Authored-By: Andronik Ordian <[email protected]> * Update ethcore/trace/src/types/error.rs Co-Authored-By: Andronik Ordian <[email protected]> * Update ethcore/types/src/engines/params.rs Co-Authored-By: Andronik Ordian <[email protected]> * Update ethcore/trace/src/types/error.rs Co-Authored-By: Andronik Ordian <[email protected]> * fix gas tier & tests * Update ethcore/evm/src/interpreter/mod.rs Co-Authored-By: Andronik Ordian <[email protected]> * Comment substack_limit test * Revert using U256 for bad jump size reporting Co-authored-by: adria0.eth <[email protected]> Co-authored-by: adria0 <[email protected]> Co-authored-by: Andronik Ordian <[email protected]>
hart36 · May 15, 2020 · 234a287 · 234a287
1 parent 1ace3c7
commit 234a287
Show file tree

Hide file tree

Showing 12 changed files with 359 additions and 30 deletions.
diff --git a/ethcore/evm/src/instructions.rs b/ethcore/evm/src/instructions.rs
@@ -41,7 +41,6 @@ macro_rules! enum_with_from_u8 {
 		}
 	};
 }
-
 enum_with_from_u8! {
 	#[doc = "Virtual machine bytecode instruction."]
 	#[repr(u8)]
@@ -321,6 +320,13 @@ enum_with_from_u8! {
 		#[doc = "Makes a log entry, 4 topics."]
 		LOG4 = 0xa4,
 
+		#[doc = "Marks the entry point to a subroutine."]
+		BEGINSUB = 0xb2,
+		#[doc = "Jumps to a defined BEGINSUB subroutine."]
+		JUMPSUB = 0xb3,
+		#[doc = "Returns from a subroutine."]
+		RETURNSUB = 0xb7,
+
 		#[doc = "create a new account with associated code"]
 		CREATE = 0xf0,
 		#[doc = "message-call into an account"]
@@ -586,6 +592,9 @@ lazy_static! {
 		arr[LOG2 as usize] = Some(InstructionInfo::new("LOG2", 4, 0, GasPriceTier::Special));
 		arr[LOG3 as usize] = Some(InstructionInfo::new("LOG3", 5, 0, GasPriceTier::Special));
 		arr[LOG4 as usize] = Some(InstructionInfo::new("LOG4", 6, 0, GasPriceTier::Special));
+		arr[BEGINSUB as usize] = Some(InstructionInfo::new("BEGINSUB", 0, 0, GasPriceTier::Base));
+		arr[JUMPSUB as usize] = Some(InstructionInfo::new("JUMPSUB", 1, 0, GasPriceTier::Low));
+		arr[RETURNSUB as usize] = Some(InstructionInfo::new("RETURNSUB", 0, 0, GasPriceTier::VeryLow));
 		arr[CREATE as usize] = Some(InstructionInfo::new("CREATE", 3, 1, GasPriceTier::Special));
 		arr[CALL as usize] = Some(InstructionInfo::new("CALL", 7, 1, GasPriceTier::Special));
 		arr[CALLCODE as usize] = Some(InstructionInfo::new("CALLCODE", 7, 1, GasPriceTier::Special));

diff --git a/ethcore/evm/src/interpreter/mod.rs b/ethcore/evm/src/interpreter/mod.rs
@@ -62,6 +62,10 @@ const TWO_POW_96: U256 = U256([0, 0x100000000, 0, 0]); //0x1 00000000 00000000 0
 const TWO_POW_224: U256 = U256([0, 0, 0, 0x100000000]); //0x1 00000000 00000000 00000000 00000000 00000000 00000000 00000000
 const TWO_POW_248: U256 = U256([0, 0, 0, 0x100000000000000]); //0x1 00000000 00000000 00000000 00000000 00000000 00000000 00000000 000000
 
+/// Maximal subroutine stack size as specified in
+/// https://eips.ethereum.org/EIPS/eip-2315.
+pub const MAX_SUB_STACK_SIZE : usize = 1024;
+
 /// Abstraction over raw vector of Bytes. Easier state management of PC.
 struct CodeReader {
 	position: ProgramCounter,
@@ -94,6 +98,8 @@ enum InstructionResult<Gas> {
 	Ok,
 	UnusedGas(Gas),
 	JumpToPosition(U256),
+	JumpToSubroutine(U256),
+	ReturnFromSubroutine(usize),
 	StopExecutionNeedsReturn {
 		/// Gas left.
 		gas: Gas,
@@ -179,8 +185,10 @@ pub struct Interpreter<Cost: CostType> {
 	do_trace: bool,
 	done: bool,
 	valid_jump_destinations: Option<Arc<BitSet>>,
+	valid_subroutine_destinations: Option<Arc<BitSet>>,
 	gasometer: Option<Gasometer<Cost>>,
 	stack: VecStack<U256>,
+	return_stack: Vec<usize>,
 	resume_output_range: Option<(U256, U256)>,
 	resume_result: Option<InstructionResult<Cost>>,
 	last_stack_ret_len: usize,
@@ -271,12 +279,14 @@ impl<Cost: CostType> Interpreter<Cost> {
 		let params = InterpreterParams::from(params);
 		let informant = informant::EvmInformant::new(depth);
 		let valid_jump_destinations = None;
+		let valid_subroutine_destinations = None;
 		let gasometer = Cost::from_u256(params.gas).ok().map(|gas| Gasometer::<Cost>::new(gas));
 		let stack = VecStack::with_capacity(schedule.stack_limit, U256::zero());
-
+		let return_stack = Vec::with_capacity(MAX_SUB_STACK_SIZE);
 		Interpreter {
 			cache, params, reader, informant,
-			valid_jump_destinations, gasometer, stack,
+			valid_jump_destinations, valid_subroutine_destinations,
+			gasometer, stack, return_stack,
 			done: false,
 			// Overridden in `step_inner` based on
 			// the result of `ext.trace_next_instruction`.
@@ -403,7 +413,7 @@ impl<Cost: CostType> Interpreter<Cost> {
 		match result {
 			InstructionResult::JumpToPosition(position) => {
 				if self.valid_jump_destinations.is_none() {
-					self.valid_jump_destinations = Some(self.cache.jump_destinations(&self.params.code_hash, &self.reader.code));
+					self.valid_jump_destinations = Some(self.cache.jump_and_sub_destinations(&self.params.code_hash, &self.reader.code).0);
 				}
 				let jump_destinations = self.valid_jump_destinations.as_ref().expect("jump_destinations are initialized on first jump; qed");
 				let pos = match self.verify_jump(position, jump_destinations) {
@@ -412,6 +422,21 @@ impl<Cost: CostType> Interpreter<Cost> {
 				};
 				self.reader.position = pos;
 			},
+			InstructionResult::JumpToSubroutine(position) => {
+				if self.valid_subroutine_destinations.is_none() {
+					self.valid_subroutine_destinations = Some(self.cache.jump_and_sub_destinations(&self.params.code_hash, &self.reader.code).1);
+				}
+				let subroutine_destinations = self.valid_subroutine_destinations.as_ref().expect("subroutine_destinations are initialized on first jump; qed");
+				let pos = match self.verify_jump(position, subroutine_destinations) {
+					Ok(x) => x,
+					Err(e) => return InterpreterResult::Done(Err(e))
+				};
+				self.return_stack.push(self.reader.position);
+				self.reader.position = pos;
+			},	
+			InstructionResult::ReturnFromSubroutine(pos) => {
+				self.reader.position = pos;
+			},
 			InstructionResult::StopExecutionNeedsReturn {gas, init_off, init_size, apply} => {
 				let mem = mem::replace(&mut self.mem, Vec::new());
 				return InterpreterResult::Done(Ok(GasLeft::NeedsReturn {
@@ -436,15 +461,17 @@ impl<Cost: CostType> Interpreter<Cost> {
 	fn verify_instruction(&self, ext: &dyn vm::Ext, instruction: Instruction, info: &InstructionInfo) -> vm::Result<()> {
 		let schedule = ext.schedule();
 
-		if (instruction == instructions::DELEGATECALL && !schedule.have_delegate_call) ||
-			(instruction == instructions::CREATE2 && !schedule.have_create2) ||
-			(instruction == instructions::STATICCALL && !schedule.have_static_call) ||
-			((instruction == instructions::RETURNDATACOPY || instruction == instructions::RETURNDATASIZE) && !schedule.have_return_data) ||
-			(instruction == instructions::REVERT && !schedule.have_revert) ||
-			((instruction == instructions::SHL || instruction == instructions::SHR || instruction == instructions::SAR) && !schedule.have_bitwise_shifting) ||
-			(instruction == instructions::EXTCODEHASH && !schedule.have_extcodehash) ||
-			(instruction == instructions::CHAINID && !schedule.have_chain_id) ||
-			(instruction == instructions::SELFBALANCE && !schedule.have_selfbalance)
+		use instructions::*;
+		if (instruction == DELEGATECALL && !schedule.have_delegate_call) ||
+			(instruction == CREATE2 && !schedule.have_create2) ||
+			(instruction == STATICCALL && !schedule.have_static_call) ||
+			((instruction == RETURNDATACOPY || instruction == RETURNDATASIZE) && !schedule.have_return_data) ||
+			(instruction == REVERT && !schedule.have_revert) ||
+			((instruction == SHL || instruction == SHR || instruction == SAR) && !schedule.have_bitwise_shifting) ||
+			(instruction == EXTCODEHASH && !schedule.have_extcodehash) ||
+			(instruction == CHAINID && !schedule.have_chain_id) ||
+			(instruction == SELFBALANCE && !schedule.have_selfbalance) ||
+			((instruction == BEGINSUB || instruction == JUMPSUB || instruction == RETURNSUB) && !schedule.have_subs)
 		{
 			return Err(vm::Error::BadInstruction {
 				instruction: instruction as u8
@@ -525,6 +552,32 @@ impl<Cost: CostType> Interpreter<Cost> {
 			instructions::JUMPDEST => {
 				// ignore
 			},
+			instructions::BEGINSUB => {
+				// ignore
+			},
+			instructions::JUMPSUB => {
+				if self.return_stack.len() >= MAX_SUB_STACK_SIZE {
+					return Err(vm::Error::OutOfSubStack {
+						wanted: 1,
+						limit: MAX_SUB_STACK_SIZE,
+					});
+				}
+				let sub_destination = self.stack.pop_back();
+				return Ok(InstructionResult::JumpToSubroutine(
+					sub_destination
+				))
+			},
+			instructions::RETURNSUB => {
+				if let Some(pos) = self.return_stack.pop() {
+					return Ok(InstructionResult::ReturnFromSubroutine(pos))
+				} else {
+					return Err(vm::Error::SubStackUnderflow {
+						wanted: 1,
+						on_stack: 0,
+					});
+				}
+
+			},
 			instructions::CREATE | instructions::CREATE2 => {
 				let endowment = self.stack.pop_back();
 				let init_off = self.stack.pop_back();
@@ -1177,6 +1230,7 @@ impl<Cost: CostType> Interpreter<Cost> {
 		if valid_jump_destinations.contains(jump) && U256::from(jump) == jump_u {
 			Ok(jump)
 		} else {
+			// Note: if jump > usize, BadJumpDestination value is trimmed
 			Err(vm::Error::BadJumpDestination {
 				destination: jump
 			})

diff --git a/ethcore/evm/src/interpreter/shared_cache.rs b/ethcore/evm/src/interpreter/shared_cache.rs
@@ -27,6 +27,7 @@ const DEFAULT_CACHE_SIZE: usize = 4 * 1024 * 1024;
 
 /// Stub for a sharing `BitSet` data in cache (reference counted)
 /// and implementing MallocSizeOf on it.
+#[derive(Clone)]
 struct Bits(Arc<BitSet>);
 
 impl MallocSizeOf for Bits {
@@ -36,9 +37,14 @@ impl MallocSizeOf for Bits {
 	}
 }
 
+#[derive(MallocSizeOf, Clone)]
+struct CacheItem {
+	jump_destination: Bits,
+	sub_entrypoint: Bits,
+}
 /// Global cache for EVM interpreter
 pub struct SharedCache {
-	jump_destinations: Mutex<MemoryLruCache<H256, Bits>>,
+	jump_destinations: Mutex<MemoryLruCache<H256, CacheItem>>,
 }
 
 impl SharedCache {
@@ -51,45 +57,54 @@ impl SharedCache {
 	}
 
 	/// Get jump destinations bitmap for a contract.
-	pub fn jump_destinations(&self, code_hash: &Option<H256>, code: &[u8]) -> Arc<BitSet> {
+	pub fn jump_and_sub_destinations(&self, code_hash: &Option<H256>, code: &[u8]) -> (Arc<BitSet>, Arc<BitSet>) {
 		if let Some(ref code_hash) = code_hash {
 			if code_hash == &KECCAK_EMPTY {
-				return Self::find_jump_destinations(code);
+				let cache_item = Self::find_jump_and_sub_destinations(code);
+				return (cache_item.jump_destination.0, cache_item.sub_entrypoint.0);
 			}
 
 			if let Some(d) = self.jump_destinations.lock().get_mut(code_hash) {
-				return d.0.clone();
+				return (d.jump_destination.0.clone(), d.sub_entrypoint.0.clone());
 			}
 		}
 
-		let d = Self::find_jump_destinations(code);
+		let d = Self::find_jump_and_sub_destinations(code);
 
 		if let Some(ref code_hash) = code_hash {
-			self.jump_destinations.lock().insert(*code_hash, Bits(d.clone()));
+			self.jump_destinations.lock().insert(*code_hash, d.clone());
 		}
 
-		d
+		(d.jump_destination.0, d.sub_entrypoint.0)
 	}
 
-	fn find_jump_destinations(code: &[u8]) -> Arc<BitSet> {
+	fn find_jump_and_sub_destinations(code: &[u8]) -> CacheItem {
 		let mut jump_dests = BitSet::with_capacity(code.len());
+		let mut sub_entrypoints = BitSet::with_capacity(code.len());
 		let mut position = 0;
 
 		while position < code.len() {
 			let instruction = Instruction::from_u8(code[position]);
 
 			if let Some(instruction) = instruction {
-				if instruction == instructions::JUMPDEST {
-					jump_dests.insert(position);
-				} else if let Some(push_bytes) = instruction.push_bytes() {
-					position += push_bytes;
+				match instruction {
+					instructions::JUMPDEST => { jump_dests.insert(position); },
+					instructions::BEGINSUB => { sub_entrypoints.insert(position); },
+					_ => {
+						if let Some(push_bytes) = instruction.push_bytes() {
+							position += push_bytes;
+						}
+					},	
 				}
 			}
 			position += 1;
 		}
 
 		jump_dests.shrink_to_fit();
-		Arc::new(jump_dests)
+		CacheItem {
+			jump_destination: Bits(Arc::new(jump_dests)),
+			sub_entrypoint: Bits(Arc::new(sub_entrypoints)),
+		}
 	}
 }
 
@@ -102,11 +117,75 @@ impl Default for SharedCache {
 #[test]
 fn test_find_jump_destinations() {
 	// given
+
+	// 0000 7F   PUSH32 0xffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffff
+	// 0021 7F   PUSH32 0xffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffff
+	// 0042 5B   JUMPDEST
+	// 0043 01   ADD
+	// 0044 60   PUSH1 0x00
+	// 0046 55   SSTORE
 	let code = hex!("7fffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffff7fffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffff5b01600055");
 
 	// when
-	let valid_jump_destinations = SharedCache::find_jump_destinations(&code);
+	let cache_item = SharedCache::find_jump_and_sub_destinations(&code);
+
+	// then
+	assert!(cache_item.jump_destination.0.iter().eq(vec![66].into_iter()));
+	assert!(cache_item.sub_entrypoint.0.is_empty());
+}
+
+#[test]
+fn test_find_jump_destinations_not_in_data_segments() {
+	// given
+
+	// 0000 60 06   PUSH1 06
+	// 0002 56      JUMP
+	// 0003 50 5B   PUSH1 0x5B	
+	// 0005 56      STOP
+	// 0006 5B      JUMPDEST
+	// 0007 60 04   PUSH1 04
+	// 0009 56      JUMP	
+	let code = hex!("600656605B565B6004");
+
+	// when
+	let cache_item = SharedCache::find_jump_and_sub_destinations(&code);
+
+	// then
+	assert!(cache_item.jump_destination.0.iter().eq(vec![6].into_iter()));
+	assert!(cache_item.sub_entrypoint.0.is_empty());
+}
+
+#[test]
+fn test_find_sub_entrypoints() {
+	// given
+
+	// see https://eips.ethereum.org/EIPS/eip-2315 for disassembly
+	let code = hex!("6800000000000000000cb300b26011b3b7b2b7");
+
+	// when
+	let cache_item = SharedCache::find_jump_and_sub_destinations(&code);
+
+	// then
+	assert!(cache_item.jump_destination.0.is_empty());
+	assert!(cache_item.sub_entrypoint.0.iter().eq(vec![12, 17].into_iter()));
+}
+
+#[test]
+fn test_find_jump_and_sub_allowing_unknown_opcodes() {
+	// precondition
+	assert!(Instruction::from_u8(0xcc) == None);
+
+	// given
+
+	// 0000 5B   JUMPDEST       
+	// 0001 CC   ???
+	// 0002 B2   BEGINSUB
+	let code = hex!("5BCCB2");
+
+	// when
+	let cache_item = SharedCache::find_jump_and_sub_destinations(&code);
 
 	// then
-	assert!(valid_jump_destinations.contains(66));
+	assert!(cache_item.jump_destination.0.iter().eq(vec![0].into_iter()));
+	assert!(cache_item.sub_entrypoint.0.iter().eq(vec![2].into_iter()));
 }