Re: [PATCH 1/2] rust: sync: add `CondVar::notify_sync`

[Alice Ryhl]

On Thu, Dec 7, 2023 at 9:22 PM Benno Lossin <benno.lossin@xxxxxxxxx> wrote:
>
> On 12/6/23 11:09, Alice Ryhl wrote:
> > Wake up another thread synchronously.
> >
> > This method behaves like `notify_one`, except that it hints to the
> > scheduler that the current thread is about to go to sleep, so it should
> > schedule the target thread on the same CPU.
> >
> > This is used by Rust Binder as a performance optimization. When sending
> > a transaction to a different process, we usually know which thread will
> > handle it, so we can schedule that thread for execution next on this
> > CPU for better cache locality.
> >
> > Signed-off-by: Alice Ryhl <aliceryhl@xxxxxxxxxx>
> > ---
> >  rust/kernel/sync/condvar.rs | 6 ++++++
> >  1 file changed, 6 insertions(+)
> >
> > diff --git a/rust/kernel/sync/condvar.rs b/rust/kernel/sync/condvar.rs
> > index b679b6f6dbeb..9861c6749ad0 100644
> > --- a/rust/kernel/sync/condvar.rs
> > +++ b/rust/kernel/sync/condvar.rs
> > @@ -155,6 +155,12 @@ fn notify(&self, count: i32, flags: u32) {
> >          };
> >      }
> >
> > +    /// Calls the kernel function to notify one thread synchronously.
> > +    pub fn notify_sync(&self) {
> > +        // SAFETY: `wait_list` points to valid memory.
> > +        unsafe { bindings::__wake_up_sync(self.wait_list.get(), bindings::TASK_NORMAL) };
>
> I took a look at the C function (i.e. __wake_up_common) and there I
> found this:
>
>     lockdep_assert_held(&wq_head->lock);
>
> So I think this function requires that the lock is held, how are you
> ensuring this?

No, we don't need to hold a lock. The call stack is:

1. __wake_up_sync
2. __wake_up_sync_key
3. __wake_up_common_lock
4. __wake_up_common

And __wake_up_common_lock will lock wq_head->lock before calling
__wake_up_common.

Alice