From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id ; Wed, 5 Mar 2003 10:18:46 -0500 Received: (majordomo@vger.kernel.org) by vger.kernel.org id ; Wed, 5 Mar 2003 10:18:46 -0500 Received: from locutus.cmf.nrl.navy.mil ([134.207.10.66]:17301 "EHLO locutus.cmf.nrl.navy.mil") by vger.kernel.org with ESMTP id ; Wed, 5 Mar 2003 10:18:45 -0500 Message-Id: <200303051528.h25FSqGi006413@locutus.cmf.nrl.navy.mil> To: "David S. Miller" cc: wa@almesberger.net, linux-kernel@vger.kernel.org Subject: Re: [PATCH][ATM] make atm (and clip) modular + try_module_get() In-reply-to: Your message of "Wed, 05 Mar 2003 06:53:41 PST." <20030305.065341.35361286.davem@redhat.com> X-url: http://www.nrl.navy.mil/CCS/people/chas/index.html X-mailer: nmh 1.0 Date: Wed, 05 Mar 2003 10:28:52 -0500 From: chas williams Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org In message <20030305.065341.35361286.davem@redhat.com>,"David S. Miller" writes : >I understand why you think you have to lock, that isn't the issue. >I'm telling you that you should be locking this crap at a much >higher level. ok, i see what you are saying now. since this is used to move the stuff queued from the old connection to the new connection i suppose clip might want something like? spin_lock_bh(&vcc->recvq.lock) skb_migrate(&vcc->recvq,©); spin_unlock_bh(&vcc->recvq.lock); that would block any more additions to the recvq (which should be sk->receieve_queue i suspect -- more on that later) while the skb are moved to copy. i am afraid i dont know much about how the clip driver operates.