Return-path: Received: from sabertooth02.qualcomm.com ([65.197.215.38]:15763 "EHLO sabertooth02.qualcomm.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753414AbaFHI27 (ORCPT ); Sun, 8 Jun 2014 04:28:59 -0400 From: Kalle Valo To: Ben Greear CC: , Subject: Re: [PATCH 1/4] ath10k: provide firmware crash info via debugfs. References: <1401904902-5842-1-git-send-email-greearb@candelatech.com> <87egz3nxdn.fsf@kamboji.qca.qualcomm.com> <5390B632.7030305@candelatech.com> <87wqcumuuw.fsf@kamboji.qca.qualcomm.com> <5391E842.4010401@candelatech.com> <874mzwkher.fsf@kamboji.qca.qualcomm.com> <539330A1.9050307@candelatech.com> Date: Sun, 8 Jun 2014 11:28:51 +0300 In-Reply-To: <539330A1.9050307@candelatech.com> (Ben Greear's message of "Sat, 7 Jun 2014 08:32:49 -0700") Message-ID: <877g4riz3w.fsf@kamboji.qca.qualcomm.com> (sfid-20140608_102911_428603_5303D27B) MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Sender: linux-wireless-owner@vger.kernel.org List-ID: Ben Greear writes: > On 06/07/2014 05:55 AM, Kalle Valo wrote: >> Ben Greear writes: >> >>>>> I'm a bit leery of adding spin-locks in the dump routine just for >>>>> this, but I can add and use a new spin-lock if you prefer. >>>> >>>> Why a new spinlock? I didn't review the locking requirements, but I >>>> would first check ar->data_lock can be used. >>> >>> I think it has to be a spin-lock because the crash dump is gathered >>> in the irq handler, so I can't use a mutex as far as I know... >>> >>> I'll work on adding such a lock today. >> >> I asked why add a _new_ spinlock as ar->data_lock is already a spinlock. > > Surely we do not want to impede traffic flow just to dump debug info? I don't see data_lock being used anywhere in hotpath, but of course I might have missed something. Anyway, if for optimisation reasons we need to introduce a new lock that should happen for a specific case in hot path. Handling firmware debug log events is not in that category. > And, it is easier to review a specific spinlock rather than use one big > global-ish lock and have to review every use of that lock for issues. Sure, it's easier for you to just add a new lock and then forget :) But not for me who would have to maintain 20 different locks for years to come. My view is that a new lock should be added on very exceptional cases and with good justifications. -- Kalle Valo