add grapher
This commit is contained in:
parent
4f0405def6
commit
ff58f88401
4
Makefile
4
Makefile
@ -2,7 +2,7 @@ DATA = $(word 1,$(wildcard ./data ../data))
|
||||
CFLAGS += -I$(DATA)
|
||||
include $(DATA)/Makefile.common
|
||||
|
||||
BINS := $(OUTDIR)/check_sanity $(OUTDIR)/make_kernel_patchfile $(OUTDIR)/apply_patchfile $(OUTDIR)/dump_range $(OUTDIR)/nm sandboxc.c
|
||||
BINS := $(OUTDIR)/check_sanity $(OUTDIR)/make_kernel_patchfile $(OUTDIR)/apply_patchfile $(OUTDIR)/dump_range $(OUTDIR)/nm $(OUTDIR)/grapher sandboxc.c
|
||||
|
||||
all: .data $(OUTDIR) $(BINS)
|
||||
|
||||
@ -24,6 +24,8 @@ $(OUTDIR)/dump_range: $(OUTDIR)/dump_range.o $(DATA)/$(OUTDIR)/libdata.a
|
||||
$(GCC) -o $@ $^
|
||||
$(OUTDIR)/nm: $(OUTDIR)/nm.o $(DATA)/$(OUTDIR)/libdata.a
|
||||
$(GCC) -o $@ $^
|
||||
$(OUTDIR)/grapher: $(OUTDIR)/grapher.o $(DATA)/$(OUTDIR)/libdata.a
|
||||
$(GXX) -o $@ $^ -O3
|
||||
|
||||
clean: .clean
|
||||
rm -f sandbox.o sandboxc.c
|
||||
|
627
grapher.cpp
Normal file
627
grapher.cpp
Normal file
@ -0,0 +1,627 @@
|
||||
#include <data/mach-o/binary.h>
|
||||
#include <stdio.h>
|
||||
#include <stdlib.h>
|
||||
#include <assert.h>
|
||||
#include <list>
|
||||
#include <unordered_map>
|
||||
#include <map>
|
||||
#include <set>
|
||||
#include <iostream>
|
||||
#include <algorithm>
|
||||
#include <vector>
|
||||
#include <mach-o/loader.h>
|
||||
#include <mach-o/nlist.h>
|
||||
#include <math.h>
|
||||
using namespace std;
|
||||
|
||||
#define STRICT 1
|
||||
|
||||
struct Edge;
|
||||
|
||||
// todo: after changing data, implement this there
|
||||
static bool b_in_vmrange(const struct binary *binary, addr_t addr) {
|
||||
CMD_ITERATE(binary->mach->hdr, cmd) {
|
||||
if(cmd->cmd == LC_SEGMENT) {
|
||||
auto s = (struct segment_command *) cmd;
|
||||
if(addr - s->vmaddr < s->vmsize) return true;
|
||||
}
|
||||
}
|
||||
return false;
|
||||
}
|
||||
|
||||
// wtf
|
||||
namespace std {
|
||||
template <typename A, typename B>
|
||||
struct hash<pair<A, B>> : public unary_function<pair<A, B>, size_t> {
|
||||
inline size_t operator()(const pair<A, B> p) const {
|
||||
hash<A> hA;
|
||||
hash<B> hB;
|
||||
return hA(p.first) ^ hB(p.second);
|
||||
}
|
||||
};
|
||||
}
|
||||
|
||||
static uint32_t signExtend(uint32_t val, int bits) {
|
||||
if(val & (1 << (bits - 1))) {
|
||||
val |= ~((1 << bits) - 1);
|
||||
}
|
||||
return val;
|
||||
}
|
||||
|
||||
enum {
|
||||
INCOMPLETE_FUNC,
|
||||
THUMB_FUNC,
|
||||
THUMB_VARARGS_FUNC,
|
||||
THUMB_ACCESSOR_FUNC
|
||||
};
|
||||
|
||||
// misnomer-- it's actually any symbol
|
||||
struct Function {
|
||||
vector<Edge *> forward, backward;
|
||||
|
||||
uint16_t *start, *end, *endOfWorld;
|
||||
addr_t startAddr;
|
||||
const char *name;
|
||||
uint32_t hash;
|
||||
|
||||
list<pair<addr_t, addr_t>> refs;
|
||||
|
||||
int type;
|
||||
|
||||
Function(uint16_t *start, uint16_t *end, uint16_t *endOfWorld, addr_t startAddr, const char *name, int type)
|
||||
: start(start), end(end), endOfWorld(endOfWorld), startAddr(startAddr), name(name), type(type) {
|
||||
if(start && !end) {
|
||||
parse();
|
||||
}
|
||||
setHash();
|
||||
}
|
||||
|
||||
void parse() {
|
||||
auto addr = startAddr & ~1;
|
||||
auto knownEnd = start;
|
||||
uint16_t *p;
|
||||
for(p = start; p + 2 <= endOfWorld && p < start + 0x400; p++, addr += 2) {
|
||||
uint32_t jumpTarget = 0;
|
||||
if((p[0] & 0xf000) == 0xd000) { // B1
|
||||
jumpTarget = signExtend(p[0] & 0xff, 8);
|
||||
p[0] = 0;
|
||||
} else if((p[0] & 0xf800) == 0xe000) { // B2
|
||||
jumpTarget = signExtend(p[0] & 0x7ff, 11);
|
||||
p[0] = 0;
|
||||
} else if((p[0] & 0xf800) == 0xf000 && ((p[1] & 0xd000) == 0x8000) && ((p[0] & 0x380) >> 7) != 7) { // B3
|
||||
jumpTarget = signExtend(((p[0] & 0x400) << 9) | ((p[1] & 0x800) << 7) | ((p[1] & 0x2000) << 4) | ((p[0] & 0x3f) << 11) | (p[1] & 0x7ff), 20);
|
||||
p[0] = p[1] = 0;
|
||||
} else if((p[0] & 0xf500) == 0xb100) { // CB[N]Z
|
||||
jumpTarget = ((p[0] & 0x200) >> 4) | ((p[0] & 0xf8) >> 3);
|
||||
} else if((p[0] & 0xf800) == 0x4800) { // LDR literal
|
||||
auto target = (uint32_t *) (p + ((addr & 2) ? 1 : 2) + 2*(p[0] & 0xff));
|
||||
if(target < (uint32_t *) endOfWorld) {
|
||||
refs.push_back(make_pair(addr, *target));
|
||||
}
|
||||
p[0] = 0;
|
||||
} else if((p[0] & 0xff7f) == 0xf85f) { // LDR literal 2
|
||||
auto target = (uint32_t *) ((uint8_t *) p + ((addr & 2) ? 2 : 4) + (p[1] & 0xfff));
|
||||
if(target < (uint32_t *) endOfWorld) {
|
||||
refs.push_back(make_pair(addr, *target));
|
||||
}
|
||||
p[0] = p[1] = 0;
|
||||
} else if((p[0] & 0xf800) == 0xf000 && (p[1] & 0xc000) == 0xc000) { // BL(X)
|
||||
// gross
|
||||
auto S = ((p[0] & 0x400) >> 10), J1 = (p[1] & 0x2000) >> 13, J2 = (p[1] & 0x800) >> 11;
|
||||
auto I1 = ~(J1 ^ S) & 1, I2 = ~(J2 ^ S) & 1;
|
||||
auto diff = ((p[0] & 0x400) >> 14) | (I1 << 23) | (I2 << 22) | ((p[0] & 0x3ff) << 12) | ((p[1] & 0x7ff) << 1);
|
||||
diff = signExtend(diff, 24);
|
||||
if(diff & 0x800000) diff |= 0xff000000;
|
||||
auto target = addr + diff + 4;
|
||||
if(p[1] & 0x1000) { // BL
|
||||
target |= 1;
|
||||
} else { // BLX
|
||||
target &= ~2;
|
||||
}
|
||||
refs.push_back(make_pair(addr, target));
|
||||
|
||||
p[0] = p[1] = 0;
|
||||
} else if(
|
||||
(type == THUMB_FUNC && p[0] == (0xbd00 | (start[0] & 0xff))) ||
|
||||
(type == THUMB_VARARGS_FUNC && p[0] == 0xb004 && p[1] == 0x4770)) { // end of function
|
||||
jumpTarget = UINT32_MAX;
|
||||
}
|
||||
|
||||
/*if(startAddr == 0x800632bd) {
|
||||
fprintf(stderr, "%x -> %x\n", startAddr + 2*(p - start), jumpTarget);
|
||||
}*/
|
||||
|
||||
if(jumpTarget) {
|
||||
if(jumpTarget != UINT32_MAX) {
|
||||
auto newEnd = p + 2 + jumpTarget;
|
||||
if(newEnd < endOfWorld && newEnd > knownEnd) knownEnd = newEnd;
|
||||
}
|
||||
if(p >= knownEnd) break;
|
||||
}
|
||||
if(((p[0] >> 13) & 0b111) == 0b111 && ((p[0] >> 11) & 0b11) != 0b00) {
|
||||
// 32-bit
|
||||
p++, addr += 2;
|
||||
}
|
||||
}
|
||||
end = p;
|
||||
}
|
||||
|
||||
void setHash() {
|
||||
hash = 0;
|
||||
for(auto p = start; p + 1 <= end && (STRICT || p < start + 10); p++) {
|
||||
hash += *p;
|
||||
}
|
||||
}
|
||||
|
||||
double predict(const Function *other) const {
|
||||
size_t myLength = end - start, hisLength = other->end - other->start;
|
||||
if(type != other->type) {
|
||||
return 0.0;
|
||||
}
|
||||
if(type == INCOMPLETE_FUNC) {
|
||||
return 0.5;
|
||||
}
|
||||
if(myLength <= 7 && hisLength <= 7) {
|
||||
return (myLength == hisLength && !memcmp(start, other->start, myLength)) ? 1.0 : 0.0;
|
||||
}
|
||||
double failed = 0;
|
||||
for(auto p = start; p < end;) {
|
||||
size_t bestBitLength = 1;
|
||||
for(auto q = other->start; q < other->end; q++) {
|
||||
size_t i = 0;
|
||||
for(i = 0; q + i < other->end && p + i < end; i++) {
|
||||
if(q[i] != p[i]) break;
|
||||
}
|
||||
if(i > bestBitLength) {
|
||||
bestBitLength = i;
|
||||
}
|
||||
}
|
||||
if(bestBitLength < 5) failed += bestBitLength;
|
||||
|
||||
p += bestBitLength;
|
||||
}
|
||||
return 1.0 - (failed / myLength);
|
||||
}
|
||||
};
|
||||
|
||||
struct Edge {
|
||||
Function *source, *dest;
|
||||
uint32_t hash;
|
||||
Edge(Function *source, Function *dest)
|
||||
: source(source), dest(dest) {
|
||||
hash = source->hash ^ ((dest->hash >> 16) | (dest->hash << 16));
|
||||
source->forward.push_back(this);
|
||||
dest->backward.push_back(this);
|
||||
}
|
||||
};
|
||||
|
||||
|
||||
struct Binary {
|
||||
struct binary binary;
|
||||
const char *filename;
|
||||
unordered_map<addr_t, Function *> funcs;
|
||||
unordered_map<uint32_t, set<Function *>> funcsByHash;
|
||||
map<string, Function *> funcsByName;
|
||||
list<Function *> funcsList;
|
||||
unordered_map<uint32_t, list<Edge *>> edgesByHash;
|
||||
|
||||
unordered_map<addr_t, const char *> reverseSymbols;
|
||||
|
||||
Binary(const char *filename)
|
||||
: filename(filename) {
|
||||
b_init(&binary);
|
||||
b_load_macho(&binary, filename);
|
||||
|
||||
doFuncs();
|
||||
doSymbols();
|
||||
}
|
||||
|
||||
void doSymbols() {
|
||||
for(uint32_t i = 0; i < binary.mach->nsyms; i++) {
|
||||
struct nlist *nl = binary.mach->symtab + i;
|
||||
if(nl->n_value && (uint32_t) nl->n_un.n_strx < binary.mach->strsize) {
|
||||
auto it = funcs.find(nl->n_value | ((nl->n_desc & N_ARM_THUMB_DEF) ? 1 : 0));
|
||||
if(it != funcs.end()) {
|
||||
setFuncName(it->second, binary.mach->strtab + nl->n_un.n_strx);
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
void setFuncName(Function *func, const char *name) {
|
||||
// not quite right
|
||||
if(func->name) {
|
||||
auto it = funcsByName.find(func->name);
|
||||
if(it != funcsByName.end() && it->second == func) funcsByName.erase(it);
|
||||
}
|
||||
|
||||
func->name = name;
|
||||
|
||||
if(name) funcsByName[name] = func;
|
||||
}
|
||||
|
||||
void setFuncHash(Function *func, uint32_t hash) {
|
||||
funcsByHash[func->hash].erase(func);
|
||||
funcsByHash[func->hash = hash].insert(func);
|
||||
}
|
||||
|
||||
Function *addFunc(uint16_t *start, uint16_t *end, addr_t addr, int type) {
|
||||
Function *&func = funcs[addr];
|
||||
if(!func) {
|
||||
func = new Function(start, end, (uint16_t *) (binary.valid_range.start + binary.valid_range.size), addr, reverseSymbols[addr], type);
|
||||
funcsList.push_back(func);
|
||||
}
|
||||
return func;
|
||||
}
|
||||
|
||||
void cut(const set<uint32_t>& cutPoints) {
|
||||
uint32_t x = 0;
|
||||
for(auto func : funcsList) {
|
||||
uint32_t hash = func->hash;
|
||||
setFuncHash(func, hash ^ x);
|
||||
if(cutPoints.find(hash) != cutPoints.end()) {
|
||||
//printf("CUT: %x\n", func->startAddr);
|
||||
x ^= (hash << 2);
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
void doFuncs() {
|
||||
auto range = b_macho_segrange(&binary, "__TEXT");
|
||||
auto pr = rangeconv(range, MUST_FIND);
|
||||
|
||||
auto start = (uint16_t *) pr.start, end = start + pr.size/2;
|
||||
auto addr = range.start;
|
||||
for(uint16_t *p = start; p + 4 <= end; p++, addr += 2) {
|
||||
if((p[0] == 0xb40f || p[0] == 0xb40c) && (p[1] & 0xff00) == 0xb500 && (p[2] & 0xff00) == 0xaf00) {
|
||||
addFunc(p, NULL, addr | 1, THUMB_VARARGS_FUNC);
|
||||
} else if((p[0] & 0xff00) == 0xb500 && (p[1] & 0xff00) == 0xaf00) {
|
||||
addFunc(p, NULL, addr | 1, THUMB_FUNC);
|
||||
} else if((p[0] & 0xf83f) == 0x6800 && p[1] == 0x4770) {
|
||||
addFunc(p, p + 2, addr | 1, THUMB_ACCESSOR_FUNC);
|
||||
}
|
||||
}
|
||||
|
||||
for(auto func : funcsList) {
|
||||
setFuncHash(func, func->hash);
|
||||
for(auto p : func->refs) {
|
||||
auto b = p.second;
|
||||
if(b_in_vmrange(&binary, b)) {
|
||||
Function *&func2 = funcs[b];
|
||||
if(!func2) {
|
||||
prange_t pr = rangeconv((range_t) {&binary, b, 4}, 0);
|
||||
if(!pr.start || b_in_vmrange(&binary, *((uint32_t *) pr.start))) {
|
||||
// quick guess
|
||||
pr.size = 0;
|
||||
}
|
||||
func2 = addFunc((uint16_t *) pr.start, (uint16_t *) (pr.start + pr.size), b, INCOMPLETE_FUNC);
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
void doEdges() {
|
||||
if(edgesByHash.size()) return;
|
||||
for(auto func : funcsList) {
|
||||
for(auto p : func->refs) {
|
||||
auto it = funcs.find(p.second);
|
||||
if(it != funcs.end()) {
|
||||
auto edge = new Edge(func, it->second);
|
||||
edgesByHash[edge->hash].push_back(edge);
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
|
||||
void identifyVtables(bool explain) {
|
||||
doEdges();
|
||||
|
||||
auto constructor = funcsByName["__ZN11OSMetaClassC2EPKcPKS_j"];
|
||||
assert(constructor);
|
||||
unordered_map<addr_t, const char *> metaClasses;
|
||||
for(auto edge : constructor->backward) {
|
||||
auto nameAddr = edge->source->refs.begin()->second;
|
||||
if(!nameAddr) continue;
|
||||
// xxx
|
||||
auto className = (const char *) rangeconv((range_t) {&binary, nameAddr, 128}, 0).start;
|
||||
if(!className) continue;
|
||||
if(edge->source->backward.size() != 1) continue;
|
||||
auto mcInstantiator = (*edge->source->backward.begin())->source;
|
||||
addr_t metaClass;
|
||||
auto it = mcInstantiator->refs.begin();
|
||||
for(it++; it != mcInstantiator->refs.end(); it++) {
|
||||
if(it->second == edge->source->startAddr) {
|
||||
auto it2 = it;
|
||||
it2--;
|
||||
metaClass = it2->second;
|
||||
goto ok;
|
||||
}
|
||||
}
|
||||
continue;
|
||||
ok:
|
||||
metaClasses[metaClass] = className;
|
||||
}
|
||||
|
||||
auto constructed = funcsByName["__ZNK11OSMetaClass19instanceConstructedEv"];
|
||||
for(auto edge : constructed->backward) {
|
||||
if(edge->source->refs.size() == 4) {
|
||||
auto it = edge->source->refs.begin();
|
||||
auto metaClass = it->second;
|
||||
it++; it++;
|
||||
auto vtable = it->second - 8;
|
||||
auto className = metaClasses[metaClass];
|
||||
if(!className) continue;
|
||||
|
||||
if(explain) printf("%x: %s\n", metaClass, className);
|
||||
|
||||
auto func = addFunc(NULL, NULL, vtable, INCOMPLETE_FUNC);
|
||||
char funcName[128];
|
||||
snprintf(funcName, sizeof(funcName), "__ZTV%zd%s", strlen(className), className);
|
||||
setFuncName(func, funcName);
|
||||
}
|
||||
}
|
||||
|
||||
//for(auto edge : func->backward) {
|
||||
}
|
||||
|
||||
void injectSymbols(const char *output) {
|
||||
// need a fresh copy that's not normalized
|
||||
struct binary binary;
|
||||
b_init(&binary);
|
||||
b_load_macho(&binary, filename);
|
||||
|
||||
char *str = binary.mach->strtab + 4;
|
||||
struct nlist *nl = binary.mach->symtab;
|
||||
for(auto p : funcsByName) {
|
||||
if(nl - binary.mach->symtab >= binary.mach->nsyms) {
|
||||
fprintf(stderr, "symbol overflow\n");
|
||||
break;
|
||||
}
|
||||
memset(nl, 0, sizeof(*nl));
|
||||
nl->n_un.n_strx = str - binary.mach->strtab;
|
||||
nl->n_value = p.second->startAddr & ~1;
|
||||
if(p.second->startAddr & 1) {
|
||||
nl->n_desc |= N_ARM_THUMB_DEF;
|
||||
}
|
||||
|
||||
auto size = p.first.size();
|
||||
if(str + size >= binary.mach->strtab + binary.mach->strsize) {
|
||||
fprintf(stderr, "string overflow at %s\n", p.first.c_str());
|
||||
abort();
|
||||
}
|
||||
strlcpy(str, p.first.data(), size + 1);
|
||||
str += size + 1;
|
||||
nl++;
|
||||
}
|
||||
|
||||
CMD_ITERATE(binary.mach->hdr, cmd) {
|
||||
if(cmd->cmd == LC_SYMTAB) {
|
||||
auto s = (struct symtab_command *) cmd;
|
||||
s->nsyms = nl - binary.mach->symtab;
|
||||
} else if(cmd->cmd == LC_DYSYMTAB) {
|
||||
auto d = (struct dysymtab_command *) cmd;
|
||||
d->iextdefsym = 0;
|
||||
d->nextdefsym = nl - binary.mach->symtab;
|
||||
}
|
||||
}
|
||||
|
||||
|
||||
b_macho_store(&binary, output);
|
||||
}
|
||||
};
|
||||
|
||||
static void doCutPoints(Binary *ba, Binary *bb) {
|
||||
set<uint32_t> cutPoints;
|
||||
for(auto p : ba->funcsByHash) {
|
||||
if(p.second.size() == 1 &&
|
||||
bb->funcsByHash[p.first].size() == 1) {
|
||||
cutPoints.insert(p.first);
|
||||
}
|
||||
}
|
||||
ba->cut(cutPoints);
|
||||
bb->cut(cutPoints);
|
||||
}
|
||||
|
||||
static list<pair<Function *, Function *>> doMatch(Binary *ba, Binary *bb) {
|
||||
doCutPoints(ba, bb);
|
||||
ba->doEdges();
|
||||
bb->doEdges();
|
||||
|
||||
// This is not the most efficient thing
|
||||
unordered_map<Function *, unordered_map<Function *, double>> xs;
|
||||
unordered_map<Edge *, unordered_map<Edge *, double>> ys;
|
||||
|
||||
for(auto a : ba->funcsList) {
|
||||
for(auto b : bb->funcsByHash[a->hash]) {
|
||||
double val;
|
||||
auto forward = a->predict(b), backward = b->predict(a);
|
||||
val = (forward + backward) / 2;
|
||||
xs[a][b] = val;
|
||||
}
|
||||
}
|
||||
|
||||
for(int iteration = 0; iteration < 6; iteration++) {
|
||||
if(0) {
|
||||
// debug
|
||||
printf("--\n");
|
||||
#define F(addr) \
|
||||
for(auto p : xs[ba->funcs[addr]]) { \
|
||||
printf("%x=%x %f\n", addr, p.first->startAddr, p.second); \
|
||||
} \
|
||||
if(1) for(auto e : ba->funcs[addr]->backward) { \
|
||||
for(auto p : ys[e]) { \
|
||||
printf("%x->%x=%x->%x %f\n", e->source->startAddr, e->dest->startAddr, p.first->source->startAddr, p.first->dest->startAddr, p.second); \
|
||||
} \
|
||||
}
|
||||
F(0x80063890)
|
||||
}
|
||||
// 5.1
|
||||
for(auto group : ba->edgesByHash) {
|
||||
for(auto edge1 : group.second) {
|
||||
for(auto edge2 : bb->edgesByHash[edge1->hash]) {
|
||||
ys[edge1][edge2] = xs[edge1->source][edge2->source] + xs[edge1->dest][edge2->dest];
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
// 5.2, tweaked to account for order
|
||||
// not perfect
|
||||
// but that needs be divided by something
|
||||
for(auto a : ba->funcsList) {
|
||||
for(auto b : bb->funcsByHash[a->hash]) {
|
||||
#define X(direction) \
|
||||
if(a->direction.size() == b->direction.size()) { \
|
||||
for(auto it = a->direction.begin(), it2 = b->direction.begin(); \
|
||||
it != a->direction.end(); \
|
||||
it++, it2++) { \
|
||||
result += ys[*it][*it2]; \
|
||||
} \
|
||||
} else { \
|
||||
double temp = 0; \
|
||||
for(auto ar : a->direction) { \
|
||||
for(auto br : b->direction) { \
|
||||
temp += ys[ar][br]; \
|
||||
} \
|
||||
} \
|
||||
result += temp / (a->direction.size() + b->direction.size()); /* not a mistake */ \
|
||||
}
|
||||
double result = 0;
|
||||
X(forward)
|
||||
X(backward)
|
||||
xs[a][b] = result;
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
list<pair<Function *, Function *>> result;
|
||||
|
||||
for(auto p : xs) {
|
||||
Function *maxFunction = NULL;
|
||||
double maxValue = -1;
|
||||
for(auto p2 : p.second) {
|
||||
if(p2.second > maxValue) {
|
||||
maxValue = p2.second;
|
||||
maxFunction = p2.first;
|
||||
}
|
||||
}
|
||||
if(maxFunction) {
|
||||
result.push_back(make_pair(p.first, maxFunction));
|
||||
}
|
||||
}
|
||||
|
||||
return result;
|
||||
}
|
||||
|
||||
static list<pair<Function *, Function *>> doMatchTrivially(Binary *ba, Binary *bb) {
|
||||
if(ba->funcsList.size() != bb->funcsList.size()) {
|
||||
fprintf(stderr, "funcs list not the same size: %d/%d\n", (int) ba->funcsList.size(), (int) bb->funcsList.size());
|
||||
}
|
||||
list<pair<Function *, Function *>> result;
|
||||
for(auto it = ba->funcsList.begin(), it2 = bb->funcsList.begin();
|
||||
it != ba->funcsList.end() && it2 != bb->funcsList.end();
|
||||
it++, it2++) {
|
||||
result.push_back(make_pair(*it, *it2));
|
||||
}
|
||||
return result;
|
||||
}
|
||||
|
||||
int main(__unused int argc, char **argv) {
|
||||
argv++;
|
||||
Binary ba(*argv++);
|
||||
while(auto arg = *argv++)
|
||||
if(!strncmp(arg, "--", 2)) {
|
||||
string mode = arg;
|
||||
if(mode == "--list") {
|
||||
printf("List of funcs:\n");
|
||||
bool refs = false;
|
||||
if(*argv && !strcmp(*argv, "--refs")) {
|
||||
refs = true;
|
||||
argv++;
|
||||
}
|
||||
for(auto func : ba.funcsList) {
|
||||
printf("%x-%x l=%ld h=%x n=%s f=%d b=%d t=%d\n", func->startAddr, (addr_t) (func->startAddr + 2*(func->end - func->start)), func->end - func->start, func->hash, func->name, (int) func->forward.size(), (int) func->backward.size(), func->type);
|
||||
if(refs) for(auto ref : func->refs) {
|
||||
printf(" r:%x->%x\n", (int) ref.first, (int) ref.second);
|
||||
}
|
||||
}
|
||||
} else if(mode == "--cut") {
|
||||
Binary bb(*argv++);
|
||||
doCutPoints(&ba, &bb);
|
||||
} else if(mode == "--byHash") {
|
||||
printf("List of funcs by hash:\n");
|
||||
|
||||
for(auto p : ba.funcsByHash) {
|
||||
printf("%d - [%08x]:", (int) p.second.size(), p.first);
|
||||
for(auto func : p.second) {
|
||||
printf(" %x", func->startAddr);
|
||||
}
|
||||
printf("\n");
|
||||
}
|
||||
|
||||
} else if(mode == "--compare") {
|
||||
Binary bb(*argv++);
|
||||
for(auto p : ba.reverseSymbols) {
|
||||
auto myAddr = p.first, otherAddr = b_sym(&bb.binary, p.second, TO_EXECUTE);
|
||||
auto first = ba.funcs[myAddr], second = bb.funcs[otherAddr];
|
||||
if(first && second) {
|
||||
double forward = first->predict(second), backward = second->predict(first);
|
||||
printf("%.32s (%08x/%08x): %f\n", p.second, myAddr, otherAddr, (forward + backward) / 2);
|
||||
}
|
||||
}
|
||||
} else if(mode == "--matchF" || mode == "--matchB" || mode == "--trivial") {
|
||||
Binary bb(*argv++);
|
||||
list<pair<Function *, Function *>> result;
|
||||
if(mode == "--matchF") {
|
||||
result = doMatch(&ba, &bb);
|
||||
} else if(mode == "--matchB") {
|
||||
for(auto p : doMatch(&bb, &ba)) result.push_back(make_pair(p.second, p.first));
|
||||
} else if(mode == "--trivial") {
|
||||
result = doMatchTrivially(&ba, &bb);
|
||||
}
|
||||
|
||||
if(!strcmp(*argv, "--audit")) {
|
||||
for(auto p : result) {
|
||||
const char *trueName = ba.reverseSymbols[p.first->startAddr];
|
||||
const char *name = bb.reverseSymbols[p.second->startAddr];
|
||||
if(name && trueName && strcmp(name, trueName)) {
|
||||
printf("Wrong: %x=%x (%s = %s)\n", p.first->startAddr, p.second->startAddr, trueName, name);
|
||||
}
|
||||
}
|
||||
argv++;
|
||||
} else if(!strcmp(*argv, "--explain")) {
|
||||
for(auto p : result) {
|
||||
auto func1 = p.first, func2 = p.second;
|
||||
printf("%08x/%08x %f/%f %s/%s\n", func1->startAddr, func2->startAddr, func1->predict(func2), func2->predict(func1), func1->name, func2->name);
|
||||
}
|
||||
argv++;
|
||||
}
|
||||
|
||||
for(auto p : result) {
|
||||
p.first->name = p.second->name;
|
||||
}
|
||||
} else if(mode == "--clear") {
|
||||
// HACK
|
||||
for(auto func : ba.funcsList) {
|
||||
if(func->name && strcmp(func->name, "__ZN11OSMetaClassC2EPKcPKS_j") && strcmp(func->name, "__ZNK11OSMetaClass19instanceConstructedEv"))
|
||||
ba.setFuncName(func, NULL);
|
||||
}
|
||||
} else if(mode == "--vt") {
|
||||
bool explain = false;
|
||||
if(*argv && !strcmp(*argv, "--explain")) {
|
||||
explain = true;
|
||||
argv++;
|
||||
}
|
||||
ba.identifyVtables(explain);
|
||||
} else {
|
||||
fprintf(stderr, "? %s\n", mode.c_str());
|
||||
abort();
|
||||
}
|
||||
} else {
|
||||
// write back
|
||||
ba.injectSymbols(arg);
|
||||
}
|
||||
|
||||
return 0;
|
||||
}
|
Loading…
Reference in New Issue
Block a user