Struct evmil::Disassembly
source · pub struct Disassembly<'a, T = ()> { /* private fields */ }
Expand description
Identifies all contiguous code blocks within the bytecode program.
Here, a block is a sequence of bytecodes terminated by either
STOP
, REVERT
, RETURN
or JUMP
. Observe that a JUMPDEST
can only appear as the first instruction of a block. In fact,
every reachable block (except the root block) begins with a
JUMPDEST
.
Implementations§
source§impl<'a, T> Disassembly<'a, T>where
T: AbstractState,
impl<'a, T> Disassembly<'a, T>where
T: AbstractState,
sourcepub fn get_block(&self, pc: usize) -> usize
pub fn get_block(&self, pc: usize) -> usize
Determine the enclosing block number for a given bytecode address.
Examples found in repository?
116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325
pub fn get_state(&self, loc: usize) -> T {
// Determine enclosing block
let bid = self.get_block(loc);
let blk = &self.blocks[bid];
let mut ctx = self.contexts[bid].clone();
let mut pc = blk.start;
// Reconstruct state
while pc < loc {
// Decode instruction at the current position
let insn = Instruction::decode(pc,&self.bytes);
// Apply the transfer function!
ctx = ctx.transfer(&insn);
// Next instruction
pc = pc + insn.length(&[]);
}
// Done
ctx
}
/// Determine the enclosing block number for a given bytecode
/// address.
pub fn get_block(&self, pc: usize) -> usize {
for i in 0..self.blocks.len() {
if self.blocks[i].encloses(pc) {
return i;
}
}
panic!("invalid bytecode address");
}
/// Determine whether a given block is currently considered
/// reachable or not. Observe the root block (`id=0`) is _always_
/// considered reachable.
pub fn is_block_reachable(&self, id: usize) -> bool {
id == 0 || self.contexts[id].is_reachable()
}
/// Read a slice of bytes from the bytecode program, padding with
/// zeros as necessary.
pub fn read_bytes(&self, start: usize, end: usize) -> Vec<u8> {
let n = self.bytes.len();
if start >= n {
vec![0; end-start]
} else if end > n {
// Determine lower potion
let mut slice = self.bytes[start..n].to_vec();
// Probably a more idiomatic way to do this?
for i in end .. n { slice.push(0); }
//
slice
} else {
// Easy case
self.bytes[start..end].to_vec()
}
}
/// Refine this disassembly to something (ideally) more precise
/// use a fixed point dataflow analysis. This destroys the
/// original disassembly.
pub fn refine<S>(self) -> Disassembly<'a,S>
where S:AbstractState+From<T> {
let mut contexts = Vec::new();
// Should be able to do this with a map?
for ctx in self.contexts {
contexts.push(S::from(ctx));
}
// Done
Disassembly{bytes: self.bytes, blocks: self.blocks, contexts}
}
/// Flattern the disassembly into a sequence of instructions.
pub fn to_vec(&self) -> Vec<Instruction> {
let mut insns = Vec::new();
let mut last = 0;
// Iterate blocks in order
for i in 0..self.blocks.len() {
let blk = &self.blocks[i];
let ctx = &self.contexts[i];
// Check for reachability
if i == 0 || ctx.is_reachable() {
// Disassemble block
self.disassemble_into(blk,&mut insns);
} else {
// Not reachable, so must be data.
let data = self.read_bytes(blk.start,blk.end);
//
insns.push(DATA(data));
}
// Update gap information
last = blk.end;
}
//
insns
}
// ================================================================
// Helpers
// ================================================================
/// Disassemble a given block into a sequence of instructions.
fn disassemble_into(&self, blk: &Block, insns: &mut Vec<Instruction>) {
let mut pc = blk.start;
// Parse the block
while pc < blk.end {
// Decode instruction at the current position
let insn = Instruction::decode(pc,&self.bytes);
// Increment PC for next instruction
pc = pc + insn.length(&[]);
//
insns.push(insn);
}
}
/// Perform a linear scan splitting out the blocks. This is an
/// over approximation of the truth, as some blocks may turn out
/// to be unreachable (e.g. they are data).
fn scan_blocks(bytes: &[u8]) -> Vec<Block> {
let mut blocks = Vec::new();
// Current position in bytecodes
let mut pc = 0;
// Identifies start of current block.
let mut start = 0;
// Parse the block
while pc < bytes.len() {
// Decode instruction at the current position
let insn = Instruction::decode(pc,&bytes);
// Increment PC for next instruction
pc = pc + insn.length(&[]);
// Check whether terminating instruction
match insn {
JUMPDEST(n) => {
// Determine whether start of this block, or next
// block.
if (pc - 1) != start {
// Start of next block
blocks.push(Block::new(start,pc-1));
start = pc - 1;
}
}
INVALID|JUMP|RETURN|REVERT|STOP => {
blocks.push(Block::new(start,pc));
start = pc;
}
_ => {}
}
}
// Append last block (if necessary)
if start != pc {
blocks.push(Block::new(start,pc));
}
// Done
blocks
}
}
impl<'a,T> Disassembly<'a,T>
where T:AbstractState+fmt::Display {
/// Apply flow analysis to refine the results of this disassembly.
pub fn build(mut self) -> Self {
let mut changed = true;
//
while changed {
// Reset indicator
changed = false;
// Iterate blocks in order
for i in 0..self.blocks.len() {
// Sanity check whether block unreachable.
if !self.is_block_reachable(i) { continue; }
// Yes, is reachable so continue.
let blk = &self.blocks[i];
let mut ctx = self.contexts[i].clone();
let mut pc = blk.start;
// println!("BLOCK (start={}, end={}): {:?}", pc, blk.end, i);
// println!("CONTEXT (pc={}): {}", pc, ctx);
// Parse the block
while pc < blk.end {
// Decode instruction at the current position
let insn = Instruction::decode(pc,&self.bytes);
// Check whether a branch is possible
if insn.can_branch() {
// Determine branch target
let target = ctx.top();
// Determine branch context
let branch_ctx = ctx.branch(target,&insn);
// Convert target into block ID.
let block_id = self.get_block(target);
// println!("Branch: target={} (block {})",target,block_id);
// println!("Before merge (pc={}): {}", pc, self.contexts[block_id]);
// Merge in updated state
changed |= self.contexts[block_id].merge(branch_ctx);
// println!("After merge (pc={}): {}", pc, self.contexts[block_id]);
}
// Apply the transfer function!
// print!("{:#08x}: {}",pc,ctx);
ctx = ctx.transfer(&insn);
// println!(" ==>\t{:?}\t==> {}",insn,ctx);
// Next instruction
pc = pc + insn.length(&[]);
}
// Merge state into following block.
if (i+1) < self.blocks.len() {
changed |= self.contexts[i+1].merge(ctx);
}
}
}
self
}
sourcepub fn is_block_reachable(&self, id: usize) -> bool
pub fn is_block_reachable(&self, id: usize) -> bool
Determine whether a given block is currently considered
reachable or not. Observe the root block (id=0
) is always
considered reachable.
Examples found in repository?
277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325
pub fn build(mut self) -> Self {
let mut changed = true;
//
while changed {
// Reset indicator
changed = false;
// Iterate blocks in order
for i in 0..self.blocks.len() {
// Sanity check whether block unreachable.
if !self.is_block_reachable(i) { continue; }
// Yes, is reachable so continue.
let blk = &self.blocks[i];
let mut ctx = self.contexts[i].clone();
let mut pc = blk.start;
// println!("BLOCK (start={}, end={}): {:?}", pc, blk.end, i);
// println!("CONTEXT (pc={}): {}", pc, ctx);
// Parse the block
while pc < blk.end {
// Decode instruction at the current position
let insn = Instruction::decode(pc,&self.bytes);
// Check whether a branch is possible
if insn.can_branch() {
// Determine branch target
let target = ctx.top();
// Determine branch context
let branch_ctx = ctx.branch(target,&insn);
// Convert target into block ID.
let block_id = self.get_block(target);
// println!("Branch: target={} (block {})",target,block_id);
// println!("Before merge (pc={}): {}", pc, self.contexts[block_id]);
// Merge in updated state
changed |= self.contexts[block_id].merge(branch_ctx);
// println!("After merge (pc={}): {}", pc, self.contexts[block_id]);
}
// Apply the transfer function!
// print!("{:#08x}: {}",pc,ctx);
ctx = ctx.transfer(&insn);
// println!(" ==>\t{:?}\t==> {}",insn,ctx);
// Next instruction
pc = pc + insn.length(&[]);
}
// Merge state into following block.
if (i+1) < self.blocks.len() {
changed |= self.contexts[i+1].merge(ctx);
}
}
}
self
}
sourcepub fn read_bytes(&self, start: usize, end: usize) -> Vec<u8>
pub fn read_bytes(&self, start: usize, end: usize) -> Vec<u8>
Read a slice of bytes from the bytecode program, padding with zeros as necessary.
Examples found in repository?
188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210
pub fn to_vec(&self) -> Vec<Instruction> {
let mut insns = Vec::new();
let mut last = 0;
// Iterate blocks in order
for i in 0..self.blocks.len() {
let blk = &self.blocks[i];
let ctx = &self.contexts[i];
// Check for reachability
if i == 0 || ctx.is_reachable() {
// Disassemble block
self.disassemble_into(blk,&mut insns);
} else {
// Not reachable, so must be data.
let data = self.read_bytes(blk.start,blk.end);
//
insns.push(DATA(data));
}
// Update gap information
last = blk.end;
}
//
insns
}
sourcepub fn refine<S>(self) -> Disassembly<'a, S>where
S: AbstractState + From<T>,
pub fn refine<S>(self) -> Disassembly<'a, S>where
S: AbstractState + From<T>,
Refine this disassembly to something (ideally) more precise use a fixed point dataflow analysis. This destroys the original disassembly.
sourcepub fn to_vec(&self) -> Vec<Instruction>
pub fn to_vec(&self) -> Vec<Instruction>
Flattern the disassembly into a sequence of instructions.