From owner-freebsd-net@FreeBSD.ORG Fri Jan 11 19:06:11 2008 Return-Path: Delivered-To: net@freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [IPv6:2001:4f8:fff6::34]) by hub.freebsd.org (Postfix) with ESMTP id 4FB6016A41A; Fri, 11 Jan 2008 19:06:11 +0000 (UTC) (envelope-from sam@errno.com) Received: from ebb.errno.com (ebb.errno.com [69.12.149.25]) by mx1.freebsd.org (Postfix) with ESMTP id DEFF513C467; Fri, 11 Jan 2008 19:06:10 +0000 (UTC) (envelope-from sam@errno.com) Received: from trouble.errno.com (trouble.errno.com [10.0.0.248]) (authenticated bits=0) by ebb.errno.com (8.13.6/8.12.6) with ESMTP id m0BISGDE011437 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-SHA bits=256 verify=NO); Fri, 11 Jan 2008 10:28:18 -0800 (PST) (envelope-from sam@errno.com) Message-ID: <4787B540.2040003@errno.com> Date: Fri, 11 Jan 2008 10:28:16 -0800 From: Sam Leffler User-Agent: Thunderbird 2.0.0.9 (X11/20071125) MIME-Version: 1.0 To: Sepherosa Ziehau References: <86ir2hznnd.fsf@ds4.des.no> <86abnpu0wv.fsf@ds4.des.no> <86abnovy4k.fsf@ds4.des.no> <86odc3dlgi.fsf@ds4.des.no> <86lk76c6t5.fsf@ds4.des.no> <864pds8idc.fsf@ds4.des.no> In-Reply-To: Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 8bit X-DCC--Metrics: ebb.errno.com; whitelist Cc: =?ISO-8859-1?Q?Dag-Erling_Sm=F8rgrav?= , kevlo@freebsd.org, current@freebsd.org, net@freebsd.org Subject: Re: if_ral regression [was seq#'s on ap frames] 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: Fri, 11 Jan 2008 19:06:11 -0000 Sepherosa Ziehau wrote: > On Jan 5, 2008 10:19 PM, Dag-Erling Smørgrav wrote: > >> "Sepherosa Ziehau" writes: >> >>> This actually brings up two things: >>> 1) I think we should ignore seq in multicast frames; this is permitted in >>> 802.11 standard. In dfly I did that, since one of our users >>> encountered a broken commercial AP which is not 802.11e but uses >>> different seq for data and beacon. >>> 2) TX sequence. I think standards only mention QSTA/nQSTA, but not >>> AP. Currently our AP uses per node TX seq, which means beacon's seq >>> is difficult to choose, at least for 2560 based ral(4), whose beacons' >>> seq need to be set by software. I saw Linksys AP uses one seq for all >>> of the frames it sends. Sam, what's you opinion on this? >>> >>> I think if STA counts ral(4)'s beacon seq, as what we do currently, >>> beacon missing will quickly happen since beacons will be discarded >>> after first several data frames. >>> >> OK, I *think* I understood most of that. Does this suggest a solution >> to you? I will try to get the wlandebug output tonight. >> > > I don't know whether you are still interested to track down the wired > problem you had experienced. > If you have time please try following patches: > > revert the old patch at your AP side and try this one > http://people.freebsd.org/~sephe/rt2560_test.diff1 > > apply following patch at you STA side > http://people.freebsd.org/~sephe/ieee80211_input.c.diff > > Best Regards, > sephe > > [just back from holiday and only now caught your question about seq#'s] The issue of seq# generation has mostly been ignored because many/most devices generate them directly. I'm aware of issues w/ management frames not getting correct seq#'s for ap mode (e.g. probe response) and have some uncommitted changes. Beacon frames should track the seq# in the bss node though there might be some trickiness w/ the bss node being rebuilt too much (we may need to propagate the current seq# as we do some other state). On the rx side I'm not sure about ignoring seq# of mcast frames; it would definitely be a WAR for broken ap implementations. FWIW I took ownership of a ral bug where AP mode tx just stopped for no apparent reason (I think it was probe response frames but can't recall). This sounds like the same thing; can you check kern/117655? Sam