Date: Thu, 19 Aug 1999 12:56:56 -0700 From: Mike Smith <mike@smith.net.au> To: Julian Elischer <julian@whistle.com> Cc: current@FreeBSD.ORG Subject: Re: HEADS UP: at_shutdown going away Message-ID: <199908191956.MAA07723@dingo.cdrom.com> In-Reply-To: Your message of "Thu, 19 Aug 1999 12:11:49 PDT." <Pine.BSF.3.95.990819120839.13522B-100000@current1.whistle.com>
next in thread | previous in thread | raw e-mail | index | archive | help
> there was external code at TRW but I don't know if it is still in use..
>
> We have uses of it here at whistle too. (that's why I wrote it..)
> but I guess I can handle them...
>
> why the change?
> If you re-aranged the arguments you'd have EVENTHANDLER_REGISTER ==
> at_shutdown_pri.
It has to do with the way that I deal with arbitrary arguments to the
called-out functions, in order to save all the crap that the caller
might have to perform with constructing an argument structure and then
passing a pointer to it.
It's easier to explain like this (backslashes elided):
#define EVENTHANDLER_DECLARE(name, type)
struct eventhandler_entry_ ## name
{
struct eventhandler_entry ee;
type eh_func;
};
#define EVENTHANDLER_INVOKE(name, args...)
{
struct eventhandler_list *el;
struct eventhandler_entry *ep;
if ((el = eventhandler_find_list(#name)) != NULL) {
for (ep = TAILQ_FIRST(&(el->el_entries));
ep != NULL;
ep = TAILQ_NEXT(ep, ee_link)) {
((struct eventhandler_entry_ ## name *)ep)->
eh_func(ep->ee_arg , ## args);
}
}
}
ie. the called-out function isn't forced to be of any particular type
other than that it can't return anything and it's first argument must
be void *. Unfortunately, the way that macro varargs works makes it
hard to put the variable arguments _first_ (and I wouldn't want to do
that since it would make calling out to a real varargs function
impossible).
> what are the selectabel (available) event queues?
That's one of the nicer features of this mechanism; you can create them
more or less on the fly. There are actually two implementations; one
optimised for speed which requires the queue to be present in the linked
objects (effectively it does the name->list translation using the
linker), the other optimised for convenience which just requires the
queue to be defined somewhere in a shared header (to get the queue entry
type into scope). This latter mechanism means that eg. I can say
EVENTHANDLER_REGISTER(pwrmgmt_suspend, foosuspend, foo_softc, 0)
from within a driver, even if the code that's going to invoke this
handler hasn't been loaded yet. By tying events to named queues rather
than to implementations, it helps soften the boundaries - in this
example the caller that eventually invokes the pwrmgmt_suspend handlers
will initially be APM, but it may later be ACPI, or something totally
different on eg. the Alpha, or an OpenFirmware system.
--
\\ The mind's the standard \\ Mike Smith
\\ of the man. \\ msmith@freebsd.org
\\ -- Joseph Merrick \\ msmith@cdrom.com
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-current" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199908191956.MAA07723>