Return-path: Received: from mail-wi0-f180.google.com ([209.85.212.180]:37689 "EHLO mail-wi0-f180.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753808AbaKXODL convert rfc822-to-8bit (ORCPT ); Mon, 24 Nov 2014 09:03:11 -0500 Received: by mail-wi0-f180.google.com with SMTP id n3so5877297wiv.7 for ; Mon, 24 Nov 2014 06:03:10 -0800 (PST) MIME-Version: 1.0 In-Reply-To: <871tos7lo0.fsf@kamboji.qca.qualcomm.com> References: <1416656922-6645-1-git-send-email-yanbol@qti.qualcomm.com> <871tos7lo0.fsf@kamboji.qca.qualcomm.com> Date: Mon, 24 Nov 2014 15:03:10 +0100 Message-ID: (sfid-20141124_150319_299757_68C7134C) Subject: Re: [PATCH] ath10k: Add the target register access and memory dump debugfs interface From: Michal Kazior To: Kalle Valo Cc: Yanbo Li , Yanbo Li , linux-wireless , "ath10k@lists.infradead.org" Content-Type: text/plain; charset=UTF-8 Sender: linux-wireless-owner@vger.kernel.org List-ID: On 24 November 2014 at 14:50, Kalle Valo wrote: > Yanbo Li writes: [...] >> + spin_lock_bh(&ar->data_lock); >> + reg_addr = ar->debug.reg_addr; >> + spin_unlock_bh(&ar->data_lock); > > [...] > >> + spin_lock_bh(&ar->data_lock); >> + ar->debug.reg_addr = reg_addr; >> + spin_unlock_bh(&ar->data_lock); > > [...] > >> + spin_lock_bh(&ar->data_lock); >> + reg_addr = ar->debug.reg_addr; >> + spin_unlock_bh(&ar->data_lock); > > [...] > >> + spin_lock_bh(&ar->data_lock); >> + reg_addr = ar->debug.reg_addr; >> + spin_unlock_bh(&ar->data_lock); > > I admit that I'm far from a locking expert, but does that make any > difference in the functionality? I imagine reg_addr could end up needing 2 instructions to be set on some weird architecture+compiler combination in which case locking is desired. With multiple access to write reg_addr you could end up reading an address that either isn't the old nor the new one. Should be harmless though and you could argue that you shouldn't access the reg_addr debugfs file simultaneously in the first place. Maybe I'm just over-exaggerating? :-) Michał