From owner-freebsd-net@FreeBSD.ORG Tue Apr 28 21:26:30 2009 Return-Path: Delivered-To: freebsd-net@freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [IPv6:2001:4f8:fff6::34]) by hub.freebsd.org (Postfix) with ESMTP id B9D86106564A for ; Tue, 28 Apr 2009 21:26:30 +0000 (UTC) (envelope-from luigi@onelab2.iet.unipi.it) Received: from onelab2.iet.unipi.it (onelab2.iet.unipi.it [131.114.9.129]) by mx1.freebsd.org (Postfix) with ESMTP id 7E1BE8FC0C for ; Tue, 28 Apr 2009 21:26:30 +0000 (UTC) (envelope-from luigi@onelab2.iet.unipi.it) Received: by onelab2.iet.unipi.it (Postfix, from userid 275) id 71D12730A4; Tue, 28 Apr 2009 23:31:41 +0200 (CEST) Date: Tue, 28 Apr 2009 23:31:41 +0200 From: Luigi Rizzo To: Andrew Snow Message-ID: <20090428213141.GC20530@onelab2.iet.unipi.it> References: <36906055-E1AE-486B-BA77-D260E0609BBB@netasq.com> <50451.74235.qm@web63901.mail.re1.yahoo.com> <20090428150739.GC8430@onelab2.iet.unipi.it> <49F7709F.1020409@modulus.org> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <49F7709F.1020409@modulus.org> User-Agent: Mutt/1.4.2.3i Cc: FreeBSD Net Subject: Re: Interrupts + Polling mode (similar to Linux's NAPI) X-BeenThere: freebsd-net@freebsd.org X-Mailman-Version: 2.1.5 Precedence: list List-Id: Networking and TCP/IP with FreeBSD List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Tue, 28 Apr 2009 21:26:31 -0000 On Wed, Apr 29, 2009 at 07:09:51AM +1000, Andrew Snow wrote: > Luigi Rizzo wrote: > >If i am not mistaken we don't have generic support for interrupt moderation > >in the kernel but that's a specific NIC feature: it works if the > >hardware supports it, and it doesn't otherwise. > > > >Of course it would be possible to modify polling to implement > >generic interrupt mitigation even without hardware support, so > >you get the best of the two worlds. > > It seems to me that you're wasting your time if you are trying to > achieve a high throughput in FreeBSD without using an Intel Pro/1000 or > 10gbe networking card. this is a very partial view of the world. the point is not getting a speed record but making the system work well on a wide variety of hardware. improving other parts of the network flow is nice and useful but it still does not address livelock and the problems that interrupt mitigation or polling are dealing with. cheers luigi > So I don't know if anyone would really miss out if generic polling > support was completely removed from the kernel and all efforts were then > placed into improving other parts of network flow in the kernel which > need more help. > > > - Andrew