From mboxrd@z Thu Jan 1 00:00:00 1970 From: jamal Subject: Re: [RFC PATCH 32/33] Add the Xen virtual network device driver. Date: Tue, 18 Jul 2006 08:39:39 -0400 Message-ID: <1153226379.5283.51.camel@jzny2> References: <20060718091807.467468000@sous-sol.org> <20060718091958.414414000@sous-sol.org> <1153218477.3038.46.camel@laptopd505.fenrus.org> Reply-To: hadi@cyberus.ca Mime-Version: 1.0 Content-Type: text/plain Content-Transfer-Encoding: 7bit Cc: Chris Wright , linux-kernel@vger.kernel.org, virtualization@lists.osdl.org, xen-devel@lists.xensource.com, Jeremy Fitzhardinge , Andi Kleen , Andrew Morton , Rusty Russell , Zachary Amsden , Ian Pratt , Christian Limpach , netdev@vger.kernel.org Return-path: Received: from mailhub.znyx.com ([208.2.156.141]:65034 "EHLO mailhub.znyx.com") by vger.kernel.org with ESMTP id S1750880AbWGRMlR (ORCPT ); Tue, 18 Jul 2006 08:41:17 -0400 To: Arjan van de Ven In-Reply-To: <1153218477.3038.46.camel@laptopd505.fenrus.org> Sender: netdev-owner@vger.kernel.org List-Id: netdev.vger.kernel.org On Tue, 2006-18-07 at 12:27 +0200, Arjan van de Ven wrote: > > Hmmm maybe it's me, but something bugs me if a NIC driver is going to > send IP level ARP packets... that just feels very very wrong and is a > blatant layering violation.... It is but the bonding driver has been setting precedence for years now on sending ARPs from a driver;-> It does make a lot of sense to put it in user space. More interesting policies may include sending more than just ARPs and once you hard-code in the kernel you loose that flexibility. > shouldn't the ifup/ifconfig scripts just > be fixed instead if this is critical behavior? > I dont think the ifup/ifconfig provide operational status (i.e link up/down) - or do they? If they can be made to invoke scripts in such a case then we are set. Note: you will get netlink events when devices are created or devices change their admin (via ifconfig) or operational (link down/up) status. [Try running "ip monitor" to see] One could write a little daemon that reacts to these specific events. The problem has been some people claiming that daemons are a bad idea from a usability perspective. Patrick has mentioned he may be working on a daemon in user space that does exactly that. The other alternative is to do the udev thing and have the kernel invoke a script whenever an event of interest happens. cheers, jamal