From: Evgeniy Polyakov Subject: Re: [take24 0/6] kevent: Generic event handling mechanism. Date: Thu, 23 Nov 2006 15:22:25 +0300 Message-ID: <20061123122225.GD20294@2ka.mipt.ru> References: <4560F07B.10608@redhat.com> <20061120082500.GA25467@2ka.mipt.ru> <4562102B.5010503@redhat.com> <20061121095302.GA15210@2ka.mipt.ru> <45633049.2000209@redhat.com> <20061121174334.GA25518@2ka.mipt.ru> <4563FD53.7030307@redhat.com> <20061122120933.GA32681@2ka.mipt.ru> <20061122121516.GA7229@2ka.mipt.ru> <4564CE00.9030904@redhat.com> Mime-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Transfer-Encoding: QUOTED-PRINTABLE Cc: David Miller , Andrew Morton , netdev , Zach Brown , Christoph Hellwig , Chase Venters , Johann Borck , linux-kernel@vger.kernel.org, Jeff Garzik , Alexander Viro Return-path: Received: from relay.2ka.mipt.ru ([194.85.82.65]:28838 "EHLO 2ka.mipt.ru") by vger.kernel.org with ESMTP id S1757358AbWKWMZW (ORCPT ); Thu, 23 Nov 2006 07:25:22 -0500 To: Ulrich Drepper Content-Disposition: inline In-Reply-To: <4564CE00.9030904@redhat.com> Sender: netdev-owner@vger.kernel.org List-Id: netdev.vger.kernel.org On Wed, Nov 22, 2006 at 02:24:00PM -0800, Ulrich Drepper (drepper@redha= t.com) wrote: > Evgeniy Polyakov wrote: > >On Wed, Nov 22, 2006 at 03:09:34PM +0300, Evgeniy Polyakov=20 > >(johnpol@2ka.mipt.ru) wrote: > >>Ok, to solve the problem in the way which should be good for both I > >>decided to implement additional syscall which will allow to mark an= y > >>event as ready and thus wake up appropriate threads. If userspace w= ill > >>request zero events to be marked as ready, syscall will just > >>interrupt/wakeup one of the listeners parked in syscall. >=20 > I'll wait for the new code drop to comment. I posted it. =20 > >Btw, what about putting aditional multiplexer into add/remove/modify > >switch? There will be logical 'ready' addon? >=20 > Is it needed? Usually this is done with a *_wait call with a timeout= of=20 > zero. That code path might have to be optimized but it should alread= y=20 > be there. It does not allow to mark events as ready. And current interfaces wake up when either timeout is zero (in this cas= e thread itself does not sleep and can process events), or when there is _new_ work - since there is no _new_ work, when thread awakened to process it was killed, kernel does not think that something is wrong. > --=20 > =E2=9E=A7 Ulrich Drepper =E2=9E=A7 Red Hat, Inc. =E2=9E=A7 444 Castro= St =E2=9E=A7 Mountain View,=20 > CA =E2=9D=96 --=20 Evgeniy Polyakov