Return-path: Received: from styx.suse.cz ([82.119.242.94]:34466 "EHLO mail.suse.cz" rhost-flags-OK-OK-OK-FAIL) by vger.kernel.org with ESMTP id S932609AbXBSUh2 (ORCPT ); Mon, 19 Feb 2007 15:37:28 -0500 Date: Mon, 19 Feb 2007 21:37:16 +0100 From: Jiri Benc To: Johannes Berg Cc: linux-wireless@vger.kernel.org, John Linville Subject: Re: [PATCH 03/10] introduce wiphy concept Message-ID: <20070219213716.183719c9@griffin.suse.cz> In-Reply-To: <20070215144257.530315000@sipsolutions.net> References: <20070215144241.847938000@sipsolutions.net> <20070215144257.530315000@sipsolutions.net> Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Sender: linux-wireless-owner@vger.kernel.org List-ID: On Thu, 15 Feb 2007 15:42:44 +0100, Johannes Berg wrote: > For the future: > > Maybe instead of having ieee80211_ptr point to a struct wireless_dev > (which can be allocated along with the netdev) we should mandate that > each wireless netdev is allocated via a new function that puts the > wireless_dev struct as the first part of the private area of the netdev, > that way getting to it is just a simple addition instead of following > a pointer... and we can also drop ieee80211_ptr from struct net_device. Currently, you need to distinguish whether some random net_device is wireless device or not at some places. At least d80211 needs that (yes, ifindex doesn't wrap often, so those are just safety checks, but - well - I'd like to be safe). How can this be effectively achieved if ieee80211_ptr is dropped? Thanks, Jiri -- Jiri Benc SUSE Labs