Re: posix message queues

[Manfred Spraul <manfred@colorfullife.com>]

Arnd Bergmann wrote:

>What worries me more than mq_attr compatibility is the conversion
>of struct sigevent, which might turn out really hard when more
>fields in there are used.
>
What is the problem for the wrappers with sigevent?

>+
>+static int get_compat_sigevent(struct sigevent *event,
>+		const struct compat_sigevent __user *u_event)
>+{
>+	if (verify_area(VERIFY_READ, u_event, sizeof(*u_event)))
>+		return -EFAULT;
>+
>+	return __get_user(event->sigev_value.sival_int,
>+			  &u_event->sigev_value.sival_int)
>+	     | __get_user(event->sigev_signo, &u_event->sigev_signo)
>+	     | __get_user(event->sigev_notify, &u_event->sigev_notify)
>+	     | __get_user(event->sigev_notify_thread_id,
>+			  &u_event->sigev_notify_thread_id);
>+}
>+
>
The dangerous case for the wrappers is mq_notify for SIGEV_THREAD: The 
kernel will read (with copy_from_user()) from the pointer 
u_event->sigev_value.sival_ptr. Fixed 32-byte buffer. The data is opaque 
- the kernel copies it back to the same process without any 
interpretation. Would that work on all archs?

--
    Manfred