From mboxrd@z Thu Jan 1 00:00:00 1970 From: David Miller Subject: Re: [PATCH net v2] bonding: fix bond_poll_controller bh_enable warning Date: Fri, 28 Aug 2015 14:13:51 -0700 (PDT) Message-ID: <20150828.141351.2056521281131696426.davem@davemloft.net> References: <658DEBCA-6BE2-4388-AFE8-58C6A71029AC@cumulusnetworks.com> <1440782540-7876-1-git-send-email-razor@blackwall.org> Mime-Version: 1.0 Content-Type: Text/Plain; charset=us-ascii Content-Transfer-Encoding: 7bit Cc: netdev@vger.kernel.org, 13806511171@163.com, shemminger@linux-foundation.org, maheshb@google.com, j.vosburgh@gmail.com, vfalico@gmail.com, gospo@cumulusnetworks.com, nikolay@cumulusnetworks.com To: razor@blackwall.org Return-path: Received: from shards.monkeyblade.net ([149.20.54.216]:55917 "EHLO shards.monkeyblade.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752349AbbH1VNw (ORCPT ); Fri, 28 Aug 2015 17:13:52 -0400 In-Reply-To: <1440782540-7876-1-git-send-email-razor@blackwall.org> Sender: netdev-owner@vger.kernel.org List-ID: From: Nikolay Aleksandrov Date: Fri, 28 Aug 2015 10:22:20 -0700 > The problem is rcu_read_unlock_bh() which triggers a warning when > irqs are disabled. ndo_poll_controller can run with bh enabled, > disabled or irqs disabled so check if that is the case and acquire > rcu_read_lock_bh only when not running with disabled irqs. I would say that having hard irqs disabled is a strict requirement, as per the debugging test in netpoll_send_skb_on_dev(): WARN_ON_ONCE(!irqs_disabled()); If you want to add the same check to netpoll_send_udp(), that's fine. But what isn't fine is adding all of this conditional locking, we want ->poll_controller() implementations to be able to depend upon the IRQ environment they execute in, otherwise every single implementation might need to have ugly conditional locking as well.