Open muniategui opened 2 weeks ago
I think we should probably do something like the following below. Then, internally, all the objects that rely on a Frida should just hold a reference to one, and we should make it all Clone
.
struct FridaSingleton {}
impl FridaSingleton {
fn new() -> Self {
unsafe { frida_sys::frida_init() };
FridaSingleton {}
}
}
impl Drop for FridaSingleton {
fn drop(&mut self) {
unsafe { frida_sys::frida_deinit() }
}
}
static THE_ONE_TRUE_FRIDA: Mutex<Option<Arc<FridaSingleton>>> = Mutex::new(None);
#[derive(Clone)]
pub struct Frida {
inner: Option<Arc<FridaSingleton>>,
}
impl Frida {
pub fn obtain() -> Self {
let mut singleton = THE_ONE_TRUE_FRIDA.lock().unwrap();
let v = singleton.get_or_insert_with(|| Arc::new(FridaSingleton::new()));
Self {
inner: Some(v.clone()),
}
}
}
impl Drop for Frida {
fn drop(&mut self) {
let Some(inner) = self.inner.take() else {
panic!("programming error!")
};
drop(inner);
let mut singleton = THE_ONE_TRUE_FRIDA.lock().unwrap();
let Some(v) = singleton.take_if(|v| Arc::strong_count(v) == 1) else {
return;
};
match Arc::try_unwrap(v) {
Ok(v) => drop(v),
Err(_v) => panic!("programming error!"),
}
}
}
This relates to https://github.com/frida/frida-rust/issues/145
If you wrap it in an Arc, doesn't that ensure drop semantics on its own. Do you really need to implement the Drop manually?
If you wrap it in an Arc, doesn't that ensure drop semantics on its own. Do you really need to implement the Drop manually?
You need to do this because static THE_ONE_TRUE_FRIDA
is holding a reference. You might think that a valid approach would be to stick a Weak
in the static rather than an Arc
, but sadly in rust the weak references will prevent the value from being dropped until all weaks are dropped :(
or anything similar such as
Causes STATUS_ACCESS_VIOLATION (using WIndows 10 as OS)
I am trying to use it in a application which use asyn so basically when the task is called 2 times it crashes. Tried in a normal code like:
And it ends in the same crash