Re: [PATCH v25 0/4] MR3504: ntdll: An implementation of SRWLOCK that closer matches Windows'.

[Zebediah Figura (＠zfigura)]

Zebediah Figura (@zfigura) commented about dlls/ntdll/sync.c:
> -
>    if (srwlock_wait(lock, 0, old))
> -
>        break;
>
> +}
> -
>
> +/* Try to acquire the list lock and start waking waiters up.
> -
> -
> -
> - This function bails in the following cases:
> -
> -
> -
> -
> - the lock become locked, we should not try to wake anyone up. And the
> -
> - new owner of the lock will be doing the waking when it releases the
> -
> - lock.
> -
> -
> - someone else acquired the list lock. In this case they will be responsible
> -
> - of waking up the waiters.
> -
> -
> - no waiters remain. No one to wake up.
> - */
>
> +static void srwlock_maybe_wake(RTL_SRWLOCK *lock, RTL_SRWLOCK old)

This is... complex, although not complex enough that I hate it.

Did you try a variant where we just always grab the spinlock when waking? Was that not performant enough?