"""probe_rsd3d_lost.py <pid>
Walk s_Resources, find lost (m_bIsLost=1) RenderSurfaceD3D entries (vfptr=0x00801a94),
sample 5 of them, print m_pSurfaceBits state.

v20-active client: m_pSurfaceBits should be NULL on lost entries (RS::Destroy ran)
v20-absent client: m_pSurfaceBits often non-NULL on lost entries (PurgeResource only)

m_pSurfaceBits is at primary+0x114. Entry IS GR-view (= primary + 0x30).
So accessed at entry + (0x114 - 0x30) = entry + 0xE4."""
import ctypes, ctypes.wintypes as wt, sys, struct

PROCESS_VM_READ = 0x10
PROCESS_QUERY_INFORMATION = 0x400
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

S_RESOURCES_M_DATA = 0x008398C4
S_RESOURCES_M_NUM  = 0x008398CC
RSD3D_VTABLE       = 0x00801a94

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)

m_data = rd_u32(h, S_RESOURCES_M_DATA)
m_num  = rd_u32(h, S_RESOURCES_M_NUM)
print(f"pid {pid}: s_Resources m_data=0x{m_data:08x} m_num={m_num}")

total = 0; lost = 0; sampled = 0
sb_null = 0; sb_nonnull = 0; nonnull_samples = []

for i in range(min(m_num, 100000)):
    entry = rd_u32(h, m_data + i * 4)
    if not entry: continue
    vt = rd_u32(h, entry)
    if vt != RSD3D_VTABLE: continue
    total += 1
    bIsLost_word = rd_u32(h, entry + 8)
    if (bIsLost_word & 0xFF) == 0: continue
    lost += 1
    # entry = GR-view; primary = entry - 0x30
    # Two buffers RS::Destroy frees: primary+0x64 and primary+0x114
    # From GR-view: +0x34 and +0xE4
    buf1 = rd_u32(h, entry + 0x34)   # primary+0x64
    buf2 = rd_u32(h, entry + 0xE4)   # primary+0x114 (m_pSurfaceBits)
    # m_pD3DSurface at primary+0x120 = entry+0xF0
    d3d  = rd_u32(h, entry + 0xF0)
    if buf1 is None or buf2 is None: continue
    if buf1 == 0 and buf2 == 0:
        sb_null += 1
    else:
        sb_nonnull += 1
        if len(nonnull_samples) < 5:
            nonnull_samples.append((entry, buf1, buf2, d3d))
    sampled += 1

print(f"RSD3D entries: total={total} lost={lost} sampled-lost={sampled}")
print(f"  both buf1+buf2 NULL (Destroy was no-op or ran):     {sb_null}  ({100*sb_null//max(sampled,1)}%)")
print(f"  at least one buf NON-NULL (would benefit from v20): {sb_nonnull}  ({100*sb_nonnull//max(sampled,1)}%)")
for entry, b1, b2, d3d in nonnull_samples:
    print(f"    entry=0x{entry:08x} buf1=0x{b1:08x} buf2=0x{b2:08x} m_pD3DSurface=0x{d3d:08x}")