[Ecls-list] Revisiting locks and signals

Matthew Mondor mm_lists at pulsar-zone.net
Mon Oct 25 05:43:56 UTC 2010


On Sun, 24 Oct 2010 21:54:33 -0400 (EDT)
Daniel Herring <dherring at tentpost.com> wrote:

> New APIs like signalfd are moving away from the random-interrupt model of 
> signals towards a more I/O friendly model.  See e.g.
> http://lwn.net/Articles/225714/

This is interesting but also appears to be OS-specific... After a quick
read I'm not even sure that the kevent referenced there has anything to
do with the BSD kqueue(2) interface (as it seems a Linux-related topic
and the mentioned author is unknown to me), but if so, I still don't
see what advantage these new interfaces have over kqueue(2), which
seems more expansible and already allow to treat signal events (as well
as many other types of kernel events) in the main event loop.

If Linux also did inherit kqueue(2) (in case those kevent patches the
article mentions really are a Linux kqueue(2) implementation), then at
least that's an interface we now would have in common between Linux and
the five major BSDs (FreeBSD, OpenBSD, NetBSD, DragonflyBSD, Darwin)...

If kevent is something else, then I got confused, as on BSD kqueue(2)
is used to obtain a file descriptor, while kevent(2) is then used in
the main loop.  I'll have to check more throughly that article and its
references next week, as the topic interests me.
-- 
Matt




More information about the ecl-devel mailing list