Files
2026-10-12 16:51:08 +08:00

111 lines
6.0 KiB
Python
Raw Permalink Blame History

This file contains ambiguous Unicode characters
This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.
# 剪口证据:列被剪调用树的全部同步原语与对外副作用(把「剪口证据只证结果相同」这条教训工具化)。
# 用法:sideeff.py <原映像 .orig.bin> <根偏移>[,<根偏移>...] [--depth N]
# 从根沿直接 bl/b(尾调用)走遍调用树(函数边界取 eh_frame),输出:
# ① 每个导入(PLT)调用:名字 + 调用点 + 所在函数(同步原语 / 文件 / 线程类单独标出)
# ② 每个间接调用点 blr / br xN(非跳表):需人工逐条解析目标
# ③ 遍历到的函数数
# 只做静态枚举,不判断「成对 / 无人等待」——那一步须按清单逐条人工证。
# Cut evidence: lists all synchronization primitives and external side effects of the cut call tree (turning the lesson "cut evidence only proves the result is the same" into a tool).
# Usage: sideeff.py <original image .orig.bin> <root offset>[,<root offset>...] [--depth N]
# Walks the whole call tree from the roots along direct bl/b (tail calls) (function boundaries from eh_frame) and outputs:
# ① every import (PLT) call: name + call site + containing function (synchronization primitive / file / thread classes are marked separately)
# ② every indirect call site blr / br xN (not a jump table): the target must be resolved manually one by one
# ③ the number of functions traversed
# Static enumeration only, with no judgment of "paired / nobody waiting" -- that step must be proven manually item by item from the list.
import struct, sys, bisect, re
from capstone import Cs, CS_ARCH_ARM64, CS_MODE_ARM
img = open(sys.argv[1], 'rb').read()
roots = [int(x, 16) for x in sys.argv[2].split(',')]
maxd = int(sys.argv[sys.argv.index('--depth') + 1]) if '--depth' in sys.argv else 64
md = Cs(CS_ARCH_ARM64, CS_MODE_ARM)
# ---- 动态段:符号表 / PLT 槽名(同 sdis.py)
# ---- Dynamic section: symbol table / PLT slot names (same as sdis.py)
m = struct.unpack_from('<I', img, 4)[0]
dyn, _, _, ehs, ehe, _ = [m + x for x in struct.unpack_from('<6i', img, m + 4)]
t = {}; o = dyn
while True:
tag, v = struct.unpack_from('<qQ', img, o); o += 16
if tag == 0: break
t.setdefault(tag, v)
SYM, STR = t[6], t[5]
def cstr(off): return img[off:img.index(b'\0', off)].decode()
got = {}
for i in range(0, t.get(2, 0), 24):
off, info, _ = struct.unpack_from('<QQq', img, t[23] + i)
got[off] = cstr(STR + struct.unpack_from('<I', img, SYM + (info >> 32) * 24)[0])
# ---- eh_frame_hdr 函数起点
# ---- eh_frame_hdr function starts
def rd(o, enc, base):
fmt, app = enc & 0xf, enc & 0x70
v, sz = {0x3: (struct.unpack_from('<I', img, o)[0], 4), 0xb: (struct.unpack_from('<i', img, o)[0], 4), 0x4: (struct.unpack_from('<Q', img, o)[0], 8)}[fmt]
if app == 0x10: v += o
if app == 0x30: v += base
return v, sz
_, ptrenc, cntenc, tblenc = img[ehs:ehs + 4]
o = ehs + 4; _, s = rd(o, ptrenc, ehs); o += s; n, s = rd(o, cntenc, ehs); o += s
starts = set()
for _ in range(n):
a, s = rd(o, tblenc, ehs); o += s; _, s = rd(o, tblenc, ehs); o += s; starts.add(a)
starts = sorted(starts)
def fbounds(a):
i = bisect.bisect_right(starts, a) - 1
return starts[i], starts[i + 1] if i + 1 < len(starts) else a + 4
# ---- PLT 桩:adrp x16 / ldr x17,[x16,#o] / add / br x17
# ---- PLT stubs: adrp x16 / ldr x17,[x16,#o] / add / br x17
plt = {}
def plt_name(a):
if a in plt: return plt[a]
w = list(md.disasm(img[a:a + 12], a)); name = None
if len(w) >= 2 and w[0].mnemonic == 'adrp' and w[0].op_str.startswith('x16') and w[1].mnemonic == 'ldr' and 'x17' in w[1].op_str:
page = int(w[0].op_str.split('#')[1], 16); off = int(re.search(r'#(0x[0-9a-f]+|\d+)\]', w[1].op_str).group(1), 0)
name = got.get(page + off)
plt[a] = name; return name
SYNC = re.compile(r'Mutex|Condition|Semaphore|Event|Barrier|ReaderWriter|Lock|Signal|Wait|Thread|Sleep|Yield|Fiber|Message|Queue|Atomic|futex|cond|sem_', re.I)
IO = re.compile(r'File|fopen|fwrite|fclose|Write|Flush|Mount|Save|Commit|Directory|Create|Delete|Rename|socket|send|connect|Notify|Event', re.I)
seen = {}; imports = []; indirect = []; parent = {}
stack = [(r, 0, None) for r in roots]
while stack:
a, d, par = stack.pop()
lo, hi = fbounds(a)
if lo in seen: continue
seen[lo] = d; parent[lo] = par
code = img[lo:hi]
for ins in md.disasm(code, lo):
mn = ins.mnemonic
if mn in ('bl', 'b') and ins.op_str.startswith('#'):
tgt = int(ins.op_str[1:], 16)
if mn == 'b' and lo <= tgt < hi: continue # 函数内跳转 | Jump within the function
nm = plt_name(tgt)
if nm: imports.append((nm, ins.address, lo)); continue
if d + 1 <= maxd: stack.append((tgt, d + 1, (lo, ins.address)))
elif mn in ('blr', 'br') and not ins.op_str.startswith('x16') and not ins.op_str.startswith('x17'):
indirect.append((ins.address, lo, mn + ' ' + ins.op_str))
print(f'根 {",".join(hex(r) for r in roots)}:遍历函数 {len(seen)} 个(深度上限 {maxd})')
def cat(nm): return '同步' if SYNC.search(nm) else ('外部I/O' if IO.search(nm) else '其它')
from collections import defaultdict
by = defaultdict(list)
for nm, at, fn in imports: by[(cat(nm), nm)].append((at, fn))
for (c, nm), lst in sorted(by.items()):
print(f'[{c}] {nm} ×{len(lst)}:' + ' '.join(f'{hex(at)}@{hex(fn)}' for at, fn in lst[:12]) + (' …' if len(lst) > 12 else ''))
print(f'间接调用 {len(indirect)} 处(br 含跳表,需人工区分):')
for at, fn, s in indirect[:200]: print(f' {hex(at)} 在 {hex(fn)}:{s}')
def path(fn):
out = []
while fn is not None and parent.get(fn) is not None:
pf, at = parent[fn]; out.append(f'{hex(pf)}@{hex(at)}'); fn = pf
return ' ← '.join(out)
if '--paths' in sys.argv:
print('同步 / 外部 I/O 导入的调用链(被调函数 ← 调用者@调用点 …,深度优先首次发现的那条):')
done = set()
for nm, at, fn in imports:
if cat(nm) == '其它' or (nm, fn) in done: continue
done.add((nm, fn)); print(f' {nm} 在 {hex(fn)} ← {path(fn)}')