57b5e43d0e
Five bugs identified and patched in retail Asheron's Call client: - v3b: palette refcount over-increment (3-byte NOP at two sites) - v5: RenderSurface PurgeResource no-op stub (vtable slot 2 thunk) - v11: two dangling-pointer crash guards (NULL-check + reorder) - v14: CEnvCell::Destroy ClipPlaneList leak (18-byte JMP to cleanup thunk) - v22: unpacker stale-pointer SEH guard (whole-function __try/__except) All five ship in leakfix.dll (117 KB, SHA d282f23c…) which is loaded by acclient.exe at process start via PE import table patching by tools/install_leakfix.py. Controlled 15-client fleet soak: unpatched control died at 26h with palette exhaustion; all 14 patched clients survived past that point and reached ≥5-day uptime. Residual ~15 MB/h growth traced to d3d9.dll's internal slab allocator (260KB surface backing buffers retained after Release). See REPORT.md §10 for the full investigation; conclusion is that it's unfixable from outside d3d9. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
112 lines
4.2 KiB
Python
112 lines
4.2 KiB
Python
"""probe_gxtri3mesh.py <pid>
|
|
Walk committed memory, find every instance of GXTri3Mesh<u16> vtable
|
|
0x007ED3B0. For each instance, read:
|
|
+0x000 vfptr (= 0x007ED3B0)
|
|
+0x26c refcount (per Release: if param_1[0x26c] == 1, free)
|
|
+0x270 aggregate-IUnknown pointer (if non-NULL, the object is delegating
|
|
AddRef/Release elsewhere — used as alternate refcount path)
|
|
Goal: characterize refcount distribution. A leak would manifest as many
|
|
instances with high refcount (engine holds many lingering refs).
|
|
"""
|
|
import ctypes, ctypes.wintypes as wt, sys, struct
|
|
|
|
PROCESS_VM_READ = 0x10
|
|
PROCESS_QUERY_INFORMATION = 0x400
|
|
MEM_COMMIT = 0x1000
|
|
MEM_PRIVATE = 0x20000
|
|
PAGE_READWRITE = 0x4
|
|
PAGE_EXECUTE_READWRITE = 0x40
|
|
|
|
k = ctypes.windll.kernel32
|
|
k.OpenProcess.argtypes = [wt.DWORD, wt.BOOL, wt.DWORD]; k.OpenProcess.restype = wt.HANDLE
|
|
k.ReadProcessMemory.argtypes = [wt.HANDLE, wt.LPCVOID, wt.LPVOID, ctypes.c_size_t, ctypes.POINTER(ctypes.c_size_t)]
|
|
k.ReadProcessMemory.restype = wt.BOOL
|
|
|
|
class MBI(ctypes.Structure):
|
|
_fields_ = [("BaseAddress", ctypes.c_void_p),
|
|
("AllocationBase", ctypes.c_void_p),
|
|
("AllocationProtect", wt.DWORD),
|
|
("RegionSize", ctypes.c_size_t),
|
|
("State", wt.DWORD),
|
|
("Protect", wt.DWORD),
|
|
("Type", wt.DWORD)]
|
|
k.VirtualQueryEx.argtypes = [wt.HANDLE, ctypes.c_void_p, ctypes.POINTER(MBI), ctypes.c_size_t]
|
|
k.VirtualQueryEx.restype = ctypes.c_size_t
|
|
|
|
VTABLE = 0x007ED3B0
|
|
|
|
def rd(h, va, n):
|
|
buf = (ctypes.c_ubyte * n)(); sz = ctypes.c_size_t(0)
|
|
if not k.ReadProcessMemory(h, va, buf, n, ctypes.byref(sz)): return None
|
|
return bytes(buf[:sz.value])
|
|
|
|
def rd_u32(h, va):
|
|
b = rd(h, va, 4); return struct.unpack('<I', b)[0] if b else None
|
|
|
|
pid = int(sys.argv[1])
|
|
h = k.OpenProcess(PROCESS_VM_READ | PROCESS_QUERY_INFORMATION, False, pid)
|
|
if not h: print(f"OpenProcess err={ctypes.get_last_error()}"); sys.exit(2)
|
|
|
|
# Pass 1: locate every vtable match.
|
|
matches = []
|
|
addr = 0
|
|
mbi = MBI()
|
|
while k.VirtualQueryEx(h, addr, ctypes.byref(mbi), ctypes.sizeof(mbi)):
|
|
base = mbi.BaseAddress or 0
|
|
size = mbi.RegionSize
|
|
if (mbi.State == MEM_COMMIT and mbi.Type == MEM_PRIVATE and
|
|
(mbi.Protect & 0xFF) in (PAGE_READWRITE, PAGE_EXECUTE_READWRITE)):
|
|
off = 0
|
|
while off < size:
|
|
chunk = min(4*1024*1024, size - off)
|
|
b = rd(h, base + off, chunk)
|
|
if b:
|
|
n_words = len(b) // 4
|
|
arr = struct.unpack(f"<{n_words}I", b[:n_words*4])
|
|
for i, v in enumerate(arr):
|
|
if v == VTABLE:
|
|
matches.append(base + off + i*4)
|
|
off += chunk
|
|
addr = base + size
|
|
if addr >= 0x80000000: break
|
|
|
|
print(f"pid {pid}: total GXTri3Mesh vtable matches = {len(matches)}")
|
|
|
|
# Pass 2: for the first N (so we don't blow time), read refcount + aggregate.
|
|
SAMPLE = min(2000, len(matches))
|
|
buckets = {'rc=0':0, 'rc=1':0, 'rc=2':0, 'rc=3':0, 'rc=4-9':0, 'rc=10-99':0, 'rc>=100':0,
|
|
'aggregate':0, 'bad':0}
|
|
max_rc = 0
|
|
sum_rc = 0
|
|
counted = 0
|
|
samples = []
|
|
for i, va in enumerate(matches[:SAMPLE]):
|
|
agg = rd_u32(h, va + 0x270)
|
|
rc = rd_u32(h, va + 0x26c)
|
|
if agg is None or rc is None:
|
|
buckets['bad'] += 1; continue
|
|
if agg != 0:
|
|
buckets['aggregate'] += 1; continue
|
|
# Plausible refcounts are 0-millions; reject only insane values
|
|
if rc > 0x40000000: # garbage / pointer-shaped value at refcount offset
|
|
buckets['bad'] += 1; continue
|
|
counted += 1
|
|
sum_rc += rc
|
|
if rc > max_rc: max_rc = rc
|
|
if rc == 0: buckets['rc=0'] += 1
|
|
elif rc == 1: buckets['rc=1'] += 1
|
|
elif rc == 2: buckets['rc=2'] += 1
|
|
elif rc == 3: buckets['rc=3'] += 1
|
|
elif rc < 10: buckets['rc=4-9'] += 1
|
|
elif rc < 100: buckets['rc=10-99'] += 1
|
|
else: buckets['rc>=100'] += 1
|
|
if rc >= 5 and len(samples) < 8:
|
|
samples.append((va, rc))
|
|
|
|
print(f"sampled {SAMPLE} (of {len(matches)}) vtable matches; standalone refcount={counted}, aggregates={buckets['aggregate']}, garbage={buckets['bad']}")
|
|
for kbl, v in buckets.items():
|
|
print(f" {kbl:10s}: {v}")
|
|
if counted:
|
|
print(f" avg refcount (standalone): {sum_rc/counted:.2f} max: {max_rc}")
|
|
for va, rc in samples:
|
|
print(f" high-refcount sample va=0x{va:08x} rc={rc}")
|