From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mail.windriver.com (mail.windriver.com [147.11.1.11]) (using TLSv1 with cipher DHE-RSA-AES256-SHA (256/256 bits)) (Client CN "mail.windriver.com", Issuer "Intel External Basic Issuing CA 3A" (not verified)) by ozlabs.org (Postfix) with ESMTPS id 6B2BDB7168 for ; Thu, 2 Sep 2010 10:48:51 +1000 (EST) Date: Thu, 2 Sep 2010 08:50:34 +0800 From: Liang Li To: Ben Hutchings Subject: Re: [v1 PATCH] ucc_geth: fix ethtool set ring param bug Message-ID: <20100902005034.GA9901@localhost> References: <1283179677-21262-1-git-send-email-liang.li@windriver.com> <1283305429-8426-1-git-send-email-liang.li@windriver.com> <1283348550.2268.7.camel@achroite.uk.solarflarecom.com> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii In-Reply-To: <1283348550.2268.7.camel@achroite.uk.solarflarecom.com> Cc: netdev@vger.kernel.org, avorontsov@ru.mvista.com, davem@davemloft.net, linuxppc-dev@ozlabs.org Reply-To: Liang Li List-Id: Linux on PowerPC Developers Mail List List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , On Wed, Sep 01, 2010 at 02:42:30PM +0100, Ben Hutchings wrote: > On Wed, 2010-09-01 at 09:43 +0800, Liang Li wrote: > > It's common sense that when we should do change to driver ring > > desc/buffer etc only after 'stop/shutdown' the device. When we > > do change while devices/driver is running, kernel oops occur: > [...] > > - ug_info->bdRingLenRx[queue] = ring->rx_pending; > > - ug_info->bdRingLenTx[queue] = ring->tx_pending; > > - > > if (netif_running(netdev)) { > > - /* FIXME: restart automatically */ > > - printk(KERN_INFO > > - "Please re-open the interface.\n"); > > + u16 rx_t; > > + u16 tx_t; > > + printk(KERN_INFO "Stopping interface %s.\n", netdev->name); > > + ucc_geth_close(netdev); > > + > > + rx_t = ug_info->bdRingLenRx[queue]; > > + tx_t = ug_info->bdRingLenTx[queue]; > > + > > + ug_info->bdRingLenRx[queue] = ring->rx_pending; > > + ug_info->bdRingLenTx[queue] = ring->tx_pending; > > + > > + printk(KERN_INFO "Reactivating interface %s.\n", netdev->name); > > + ret = ucc_geth_open(netdev); > > + if (ret) { > > + printk(KERN_WARNING "uec_set_ringparam: set ring param for running" > > + " interface %s failed. Please try to make the interface " > > + " down, then try again.\n", netdev->name); > > + ug_info->bdRingLenRx[queue] = rx_t; > > + ug_info->bdRingLenTx[queue] = tx_t; > > + } > [...] > > Bringing the interface down will call ucc_geth_close(), which will try > to free resources that have not been allocated! Sorry, I did not understand you on this point. There is no ucc_geth_close when 'open fail'. What you mean here exactly? > > If you cannot roll back a failed change then at least use dev_close() > and dev_open() rather than calling ucc_geth_{close,open}() directly, so > that the interface state is updated correctly. Or just refuse to make > the change if the interface is up. That make things more simply but I do not think that is necessary. Since there is no such restriction exist in most NIC drivers. :) Actually I did not see the 'fail of reopen' case. So I assume you may witnessed similar 'open fail' case in some rare cases. May you please give more input on this then I can 'make re-open safer' here. Thanks, -Liang Li > > Ben. > > -- > Ben Hutchings, Senior Software Engineer, Solarflare Communications > Not speaking for my employer; that's the marketing department's job. > They asked us to note that Solarflare product names are trademarked.