Repository navigation
Windows: attach fails with TransportError "timeout was reached" when frida-helper is slow (GDBus 25 s default timeout) #1272
Description
Summary
On Windows, device.attach() to a 32-bit program fails with TransportError: timeout was reached when frida-helper is slow to answer. The failure comes at exactly 25 s, which is GLib's default D-Bus call timeout. throw_dbus_error maps the GIO timeout to Frida.Error.TRANSPORT, so a slow helper reads as a broken transport instead of a timeout.
We hit this on a heavily loaded machine: an attach to a freshly spawned 32-bit program took 42.7 s and failed with this error. On a quiet machine the same attach passes 20 of 20 times. The repro below makes it deterministic by suspending the helper.
Environment
- frida 17.24.3 (Python bindings), Python 3.12.4
- Windows 11 x64
- Target: a 32-bit program (SysWOW64), injected through frida-helper
Repro
The script starts the helper with one attach and detach, suspends only the helper processes that are its own children for 40 s, then spawns a 32-bit program and attaches to it.
import ctypes, os, subprocess, sys, threading, time
import frida
PROGRAM = [r"C:\Windows\SysWOW64\ping.exe", "-n", "60", "127.0.0.1"]
HOLD = 40
ntdll = ctypes.WinDLL("ntdll")
kernel32 = ctypes.WinDLL("kernel32")
def own_helpers():
q = ("Get-CimInstance Win32_Process | Where-Object { $_.ParentProcessId -eq %d "
"-and $_.Name -like 'frida-helper*' } | ForEach-Object { $_.ProcessId }" % os.getpid())
out = subprocess.run(["powershell", "-NoProfile", "-Command", q],
capture_output=True, text=True).stdout
return [int(x) for x in out.split() if x.isdigit()]
print("frida", frida.__version__, "python", sys.version.split()[0])
device = frida.get_local_device()
warm = device.spawn(PROGRAM) # starts the helper; it stays for the next attach
device.attach(warm).detach()
device.kill(warm)
handles = []
for pid in own_helpers():
h = kernel32.OpenProcess(0x0800, False, pid) # PROCESS_SUSPEND_RESUME
ntdll.NtSuspendProcess(h)
handles.append(h)
timer = threading.Timer(HOLD, lambda: [ntdll.NtResumeProcess(h) for h in handles])
timer.start()
t0 = time.time()
pid = device.spawn(PROGRAM)
try:
device.attach(pid).detach()
print("attached at %.1f s" % (time.time() - t0))
except Exception as e:
print("attach failed at %.1f s: %s: %s" % (time.time() - t0, type(e).__name__, e))
finally:
timer.join()
for h in handles:
kernel32.CloseHandle(h)
device.kill(pid)Observed
attach failed at 25.0 s: TransportError: timeout was reached
Expected
The attach waits for the helper, bounded by the caller's cancellable, or, if a limit applies, fails with TimedOutError, so a caller can tell a slow helper from a broken transport and retry.
Related
- Failed to attach: timeout was reached frida#1576 and Failed to spawn: timeout was reached #445 report the same message, but for Android or USB devices. This one is the local Windows helper path.
Activity