Official Go implementation of the Ethereum protocol
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
 
 
 
 
 
 
go-ethereum/core/vm/eof_validation.go

255 lines
10 KiB

// Copyright 2024 The go-ethereum Authors
// This file is part of the go-ethereum library.
//
// The go-ethereum library is free software: you can redistribute it and/or modify
// it under the terms of the GNU Lesser General Public License as published by
// the Free Software Foundation, either version 3 of the License, or
// (at your option) any later version.
//
// The go-ethereum library is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
// GNU Lesser General Public License for more details.
//
// You should have received a copy of the GNU Lesser General Public License
// along with the go-ethereum library. If not, see <http://www.gnu.org/licenses/>.
package vm
import (
"errors"
"fmt"
"io"
)
// Below are all possible errors that can occur during validation of
// EOF containers.
var (
errInvalidMagic = errors.New("invalid magic")
errUndefinedInstruction = errors.New("undefined instruction")
errTruncatedImmediate = errors.New("truncated immediate")
errInvalidSectionArgument = errors.New("invalid section argument")
errInvalidCallArgument = errors.New("callf into non-returning section")
errInvalidDataloadNArgument = errors.New("invalid dataloadN argument")
errInvalidJumpDest = errors.New("invalid jump destination")
errInvalidBackwardJump = errors.New("invalid backward jump")
errInvalidOutputs = errors.New("invalid number of outputs")
errInvalidMaxStackHeight = errors.New("invalid max stack height")
errInvalidCodeTermination = errors.New("invalid code termination")
errEOFCreateWithTruncatedSection = errors.New("eofcreate with truncated section")
errOrphanedSubcontainer = errors.New("subcontainer not referenced at all")
errIncompatibleContainerKind = errors.New("incompatible container kind")
errStopAndReturnContract = errors.New("Stop/Return and Returncontract in the same code section")
errStopInInitCode = errors.New("initcode contains a RETURN or STOP opcode")
errTruncatedTopLevelContainer = errors.New("truncated top level container")
errUnreachableCode = errors.New("unreachable code")
errInvalidNonReturningFlag = errors.New("invalid non-returning flag, bad RETF")
errInvalidVersion = errors.New("invalid version")
errMissingTypeHeader = errors.New("missing type header")
errInvalidTypeSize = errors.New("invalid type section size")
errMissingCodeHeader = errors.New("missing code header")
errInvalidCodeSize = errors.New("invalid code size")
errInvalidContainerSectionSize = errors.New("invalid container section size")
errMissingDataHeader = errors.New("missing data header")
errMissingTerminator = errors.New("missing header terminator")
errTooManyInputs = errors.New("invalid type content, too many inputs")
errTooManyOutputs = errors.New("invalid type content, too many outputs")
errInvalidSection0Type = errors.New("invalid section 0 type, input and output should be zero and non-returning (0x80)")
errTooLargeMaxStackHeight = errors.New("invalid type content, max stack height exceeds limit")
errInvalidContainerSize = errors.New("invalid container size")
)
const (
notRefByEither = iota
refByReturnContract
refByEOFCreate
)
type validationResult struct {
visitedCode map[int]struct{}
visitedSubContainers map[int]int
isInitCode bool
isRuntime bool
}
// validateCode validates the code parameter against the EOF v1 validity requirements.
func validateCode(code []byte, section int, container *Container, jt *JumpTable, isInitCode bool) (*validationResult, error) {
var (
i = 0
// Tracks the number of actual instructions in the code (e.g.
// non-immediate values). This is used at the end to determine
// if each instruction is reachable.
count = 0
op OpCode
analysis bitvec
visitedCode map[int]struct{}
visitedSubcontainers map[int]int
hasReturnContract bool
hasStop bool
)
// This loop visits every single instruction and verifies:
// * if the instruction is valid for the given jump table.
// * if the instruction has an immediate value, it is not truncated.
// * if performing a relative jump, all jump destinations are valid.
// * if changing code sections, the new code section index is valid and
// will not cause a stack overflow.
for i < len(code) {
count++
op = OpCode(code[i])
if jt[op].undefined {
return nil, fmt.Errorf("%w: op %s, pos %d", errUndefinedInstruction, op, i)
}
size := int(immediates[op])
if size != 0 && len(code) <= i+size {
return nil, fmt.Errorf("%w: op %s, pos %d", errTruncatedImmediate, op, i)
}
switch op {
case RJUMP, RJUMPI:
if err := checkDest(code, &analysis, i+1, i+3, len(code)); err != nil {
return nil, err
}
case RJUMPV:
max_size := int(code[i+1])
length := max_size + 1
if len(code) <= i+length {
return nil, fmt.Errorf("%w: jump table truncated, op %s, pos %d", errTruncatedImmediate, op, i)
}
offset := i + 2
for j := 0; j < length; j++ {
if err := checkDest(code, &analysis, offset+j*2, offset+(length*2), len(code)); err != nil {
return nil, err
}
}
i += 2 * max_size
case CALLF:
arg, _ := parseUint16(code[i+1:])
if arg >= len(container.types) {
return nil, fmt.Errorf("%w: arg %d, last %d, pos %d", errInvalidSectionArgument, arg, len(container.types), i)
}
if container.types[arg].outputs == 0x80 {
return nil, fmt.Errorf("%w: section %v", errInvalidCallArgument, arg)
}
if visitedCode == nil {
visitedCode = make(map[int]struct{})
}
visitedCode[arg] = struct{}{}
case JUMPF:
arg, _ := parseUint16(code[i+1:])
if arg >= len(container.types) {
return nil, fmt.Errorf("%w: arg %d, last %d, pos %d", errInvalidSectionArgument, arg, len(container.types), i)
}
if container.types[arg].outputs != 0x80 && container.types[arg].outputs > container.types[section].outputs {
return nil, fmt.Errorf("%w: arg %d, last %d, pos %d", errInvalidOutputs, arg, len(container.types), i)
}
if visitedCode == nil {
visitedCode = make(map[int]struct{})
}
visitedCode[arg] = struct{}{}
case DATALOADN:
arg, _ := parseUint16(code[i+1:])
// TODO why are we checking this? We should just pad
if arg+32 > len(container.data) {
return nil, fmt.Errorf("%w: arg %d, last %d, pos %d", errInvalidDataloadNArgument, arg, len(container.data), i)
}
case RETURNCONTRACT:
if !isInitCode {
return nil, errIncompatibleContainerKind
}
arg := int(code[i+1])
if arg >= len(container.subContainers) {
return nil, fmt.Errorf("%w: arg %d, last %d, pos %d", errUnreachableCode, arg, len(container.subContainers), i)
}
if visitedSubcontainers == nil {
visitedSubcontainers = make(map[int]int)
}
// We need to store per subcontainer how it was referenced
if v, ok := visitedSubcontainers[arg]; ok && v != refByReturnContract {
return nil, fmt.Errorf("section already referenced, arg :%d", arg)
}
if hasStop {
return nil, errStopAndReturnContract
}
hasReturnContract = true
visitedSubcontainers[arg] = refByReturnContract
case EOFCREATE:
arg := int(code[i+1])
if arg >= len(container.subContainers) {
return nil, fmt.Errorf("%w: arg %d, last %d, pos %d", errUnreachableCode, arg, len(container.subContainers), i)
}
if ct := container.subContainers[arg]; len(ct.data) != ct.dataSize {
return nil, fmt.Errorf("%w: container %d, have %d, claimed %d, pos %d", errEOFCreateWithTruncatedSection, arg, len(ct.data), ct.dataSize, i)
}
if visitedSubcontainers == nil {
visitedSubcontainers = make(map[int]int)
}
// We need to store per subcontainer how it was referenced
if v, ok := visitedSubcontainers[arg]; ok && v != refByEOFCreate {
return nil, fmt.Errorf("section already referenced, arg :%d", arg)
}
visitedSubcontainers[arg] = refByEOFCreate
case STOP, RETURN:
if isInitCode {
return nil, errStopInInitCode
}
if hasReturnContract {
return nil, errStopAndReturnContract
}
hasStop = true
}
i += size + 1
}
// Code sections may not "fall through" and require proper termination.
// Therefore, the last instruction must be considered terminal or RJUMP.
if !terminals[op] && op != RJUMP {
return nil, fmt.Errorf("%w: end with %s, pos %d", errInvalidCodeTermination, op, i)
}
if paths, err := validateControlFlow(code, section, container.types, jt); err != nil {
return nil, err
} else if paths != count {
// TODO(matt): return actual position of unreachable code
return nil, errUnreachableCode
}
return &validationResult{
visitedCode: visitedCode,
visitedSubContainers: visitedSubcontainers,
isInitCode: hasReturnContract,
isRuntime: hasStop,
}, nil
}
// checkDest parses a relative offset at code[0:2] and checks if it is a valid jump destination.
func checkDest(code []byte, analysis *bitvec, imm, from, length int) error {
if len(code) < imm+2 {
return io.ErrUnexpectedEOF
}
if analysis != nil && *analysis == nil {
*analysis = eofCodeBitmap(code)
}
offset := parseInt16(code[imm:])
dest := from + offset
if dest < 0 || dest >= length {
return fmt.Errorf("%w: out-of-bounds offset: offset %d, dest %d, pos %d", errInvalidJumpDest, offset, dest, imm)
}
if !analysis.codeSegment(uint64(dest)) {
return fmt.Errorf("%w: offset into immediate: offset %d, dest %d, pos %d", errInvalidJumpDest, offset, dest, imm)
}
return nil
}
//// disasm is a helper utility to show a sequence of comma-separated operations,
//// with immediates shown inline,
//// e.g: PUSH1(0x00),EOFCREATE(0x00),
//func disasm(code []byte) string {
// var ops []string
// for i := 0; i < len(code); i++ {
// var op string
// if args := immediates[code[i]]; args > 0 {
// op = fmt.Sprintf("%v(%#x)", OpCode(code[i]).String(), code[i+1:i+1+int(args)])
// i += int(args)
// } else {
// op = OpCode(code[i]).String()
// }
// ops = append(ops, op)
// }
// return strings.Join(ops, ",")
//}