Return-path: Received: from mail.w1.fi ([212.71.239.96]:50422 "EHLO li674-96.members.linode.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753072AbaF2SIg (ORCPT ); Sun, 29 Jun 2014 14:08:36 -0400 Date: Sun, 29 Jun 2014 21:08:33 +0300 From: Jouni Malinen To: Arik Nemtsov Cc: "linux-wireless@vger.kernel.org" , Johannes Berg Subject: Re: [PATCH 05/10] mac80211: use TDLS initiator in tdls_mgmt operations Message-ID: <20140629180833.GA15371@w1.fi> (sfid-20140629_201001_500276_B8A8796D) References: <1402496307-32522-1-git-send-email-arik@wizery.com> <1402496307-32522-5-git-send-email-arik@wizery.com> <20140629104318.GA32230@w1.fi> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii In-Reply-To: Sender: linux-wireless-owner@vger.kernel.org List-ID: On Sun, Jun 29, 2014 at 06:23:01PM +0300, Arik Nemtsov wrote: > Well I have to say I didn't really consider mac80211 TDLS > functionality "ready for prime time" when I've discovered this bug. > It's something pretty basic. > The breakage of older userspace was semi-deliberate here. It may have lacked some functionality, but it is being deployed at least in number of test setups.. Any such breakage of already exposed kernel functionality needs to really be clearly identified in the commit message, so that it is obvious to whoever may be reviewing the changes. > But your logic is sound. If you consider this important, we can fix this. I'm more concerned about non-mac80211 drivers since there are number of those with TDLS support in deployed devices. However, ath9k enabled TDLS support already in 2011 and as far as I know, it was working more or less fine for single channel use cases since that point in time (even if some of the teardown operations may have failed based on these new patches and frames getting delivered through incorrect path, etc.). While I may not be able to point to any specific deployment of mac80211-based TDLS implementation, I don't think that it is acceptable to break functionality that has been enabled for multiple years.. So yes, I think that these patches will need to be fixed or reverted to avoid such regressions. -- Jouni Malinen PGP id EFC895FA