Re: Kevent POSIX timers support.

[Ulrich Drepper <drepper@redhat.com>]

Evgeniy Polyakov wrote:
> +static int posix_kevent_init(void)
> +{
> +	struct kevent_callbacks tc = {
> +		.callback = &posix_kevent_callback,
> +		.enqueue = &posix_kevent_enqueue,
> +		.dequeue = &posix_kevent_dequeue};

How do we prevent that somebody tries to register a POSIX timer event 
source with kevent_ctl(KEVENT_CTL_ADD)?  This should only be possible 
from sys_timer_create and nowhere else.

Can you add a parameter to kevent_enqueue indicating this is a call from 
inside the kernel and then ignore certain enqueue callbacks?


> @@ -343,23 +439,27 @@ static int posix_timer_fn(struct hrtimer
>  
>  	timr = container_of(timer, struct k_itimer, it.real.timer);
>  	spin_lock_irqsave(&timr->it_lock, flags);
> +	
> +	if (timr->it_sigev_notify & SIGEV_KEVENT) {
> +		kevent_storage_ready(&timr->st, NULL, KEVENT_MASK_ALL);
> +	} else {

We need to pass the data in the sigev_value meember of the struct 
sigevent structure passed to timer_create to the caller.  I don't see it 
being done here nor when the timer is created.  Do I miss something? 
The sigev_value value should be stored in the user/ptr member of struct 
ukevent.


> +		if (event.sigev_notify & SIGEV_KEVENT) {

Don't use a bit.  It makes no sense to combine SIGEV_SIGNAL with 
SIGEV_KEVENT etc.  Only SIGEV_THREAD_ID is a special case.

Just define SIGEV_KEVENT to 3 and replace the tests like the one cited 
above with

   if (timr->it_sigev_notify == SIGEV_KEVENT)

-- 
➧ Ulrich Drepper ➧ Red Hat, Inc. ➧ 444 Castro St ➧ Mountain View, CA ❖