From patchwork Thu Sep 2 15:45:54 2010 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Jean Delvare X-Patchwork-Id: 63499 X-Patchwork-Delegate: davem@davemloft.net Return-Path: X-Original-To: patchwork-incoming@ozlabs.org Delivered-To: patchwork-incoming@ozlabs.org Received: from vger.kernel.org (vger.kernel.org [209.132.180.67]) by ozlabs.org (Postfix) with ESMTP id 5B490B7192 for ; Fri, 3 Sep 2010 01:46:05 +1000 (EST) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1752554Ab0IBPp5 (ORCPT ); Thu, 2 Sep 2010 11:45:57 -0400 Received: from cantor.suse.de ([195.135.220.2]:52767 "EHLO mx1.suse.de" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752433Ab0IBPp5 (ORCPT ); Thu, 2 Sep 2010 11:45:57 -0400 Received: from relay1.suse.de (charybdis-ext.suse.de [195.135.221.2]) (using TLSv1 with cipher DHE-RSA-AES256-SHA (256/256 bits)) (No client certificate requested) by mx1.suse.de (Postfix) with ESMTP id 46B2B8FEA2; Thu, 2 Sep 2010 17:45:56 +0200 (CEST) From: Jean Delvare Organization: Suse Linux To: Eric Dumazet , Jay Vosburgh Subject: [PATCH v2] bonding: Fix jiffies overflow problems (again) Date: Thu, 2 Sep 2010 17:45:54 +0200 User-Agent: KMail/1.12.4 (Linux/2.6.32.13-0.5-pae; KDE/4.3.5; i686; ; ) Cc: bonding-devel@lists.sourceforge.net, netdev@vger.kernel.org, Jiri Bohac References: <201009021619.46206.jdelvare@suse.de> <1283438186.2454.856.camel@edumazet-laptop> In-Reply-To: <1283438186.2454.856.camel@edumazet-laptop> MIME-Version: 1.0 Message-Id: <201009021745.55038.jdelvare@suse.de> Sender: netdev-owner@vger.kernel.org Precedence: bulk List-ID: X-Mailing-List: netdev@vger.kernel.org From: Jiri Bohac The time_before_eq()/time_after_eq() functions operate on unsigned long and only work if the difference between the two compared values is smaller than half the range of unsigned long (31 bits on i386). Some of the variables (slave->jiffies, dev->trans_start, dev->last_rx) used by bonding store a copy of jiffies and may not be updated for a long time. With HZ=1000, time_before_eq()/time_after_eq() will start giving bad results after ~25 days. jiffies will never be before slave->jiffies, dev->trans_start, dev->last_rx by more than possibly a couple ticks caused by preemption of this code. This allows us to detect/prevent these overflows by replacing time_before_eq()/time_after_eq() with time_in_range(). Signed-off-by: Jiri Bohac Signed-off-by: Jean Delvare --- drivers/net/bonding/bond_main.c | 56 +++++++++++++++++++++++++++------------ 1 file changed, 39 insertions(+), 17 deletions(-) Changes in V2: * Cache dev_trans_start() result as this function can be slow (as suggested by Eric Dumazet.) --- linux-2.6.36-rc3.orig/drivers/net/bonding/bond_main.c 2010-08-17 09:39:12.000000000 +0200 +++ linux-2.6.36-rc3/drivers/net/bonding/bond_main.c 2010-09-02 17:06:48.000000000 +0200 @@ -2797,9 +2797,15 @@ void bond_loadbalance_arp_mon(struct wor * so it can wait */ bond_for_each_slave(bond, slave, i) { + unsigned long trans_start = dev_trans_start(slave->dev); + if (slave->link != BOND_LINK_UP) { - if (time_before_eq(jiffies, dev_trans_start(slave->dev) + delta_in_ticks) && - time_before_eq(jiffies, slave->dev->last_rx + delta_in_ticks)) { + if (time_in_range(jiffies, + trans_start - delta_in_ticks, + trans_start + delta_in_ticks) && + time_in_range(jiffies, + slave->dev->last_rx - delta_in_ticks, + slave->dev->last_rx + delta_in_ticks)) { slave->link = BOND_LINK_UP; slave->state = BOND_STATE_ACTIVE; @@ -2827,8 +2833,12 @@ void bond_loadbalance_arp_mon(struct wor * when the source ip is 0, so don't take the link down * if we don't know our ip yet */ - if (time_after_eq(jiffies, dev_trans_start(slave->dev) + 2*delta_in_ticks) || - (time_after_eq(jiffies, slave->dev->last_rx + 2*delta_in_ticks))) { + if (!time_in_range(jiffies, + trans_start - delta_in_ticks, + trans_start + 2 * delta_in_ticks) || + !time_in_range(jiffies, + slave->dev->last_rx - delta_in_ticks, + slave->dev->last_rx + 2 * delta_in_ticks)) { slave->link = BOND_LINK_DOWN; slave->state = BOND_STATE_BACKUP; @@ -2883,13 +2893,16 @@ static int bond_ab_arp_inspect(struct bo { struct slave *slave; int i, commit = 0; + unsigned long trans_start; bond_for_each_slave(bond, slave, i) { slave->new_link = BOND_LINK_NOCHANGE; if (slave->link != BOND_LINK_UP) { - if (time_before_eq(jiffies, slave_last_rx(bond, slave) + - delta_in_ticks)) { + if (time_in_range(jiffies, + slave_last_rx(bond, slave) - delta_in_ticks, + slave_last_rx(bond, slave) + delta_in_ticks)) { + slave->new_link = BOND_LINK_UP; commit++; } @@ -2902,8 +2915,9 @@ static int bond_ab_arp_inspect(struct bo * active. This avoids bouncing, as the last receive * times need a full ARP monitor cycle to be updated. */ - if (!time_after_eq(jiffies, slave->jiffies + - 2 * delta_in_ticks)) + if (time_in_range(jiffies, + slave->jiffies - delta_in_ticks, + slave->jiffies + 2 * delta_in_ticks)) continue; /* @@ -2921,8 +2935,10 @@ static int bond_ab_arp_inspect(struct bo */ if (slave->state == BOND_STATE_BACKUP && !bond->current_arp_slave && - time_after(jiffies, slave_last_rx(bond, slave) + - 3 * delta_in_ticks)) { + !time_in_range(jiffies, + slave_last_rx(bond, slave) - delta_in_ticks, + slave_last_rx(bond, slave) + 3 * delta_in_ticks)) { + slave->new_link = BOND_LINK_DOWN; commit++; } @@ -2933,11 +2949,15 @@ static int bond_ab_arp_inspect(struct bo * - (more than 2*delta since receive AND * the bond has an IP address) */ + trans_start = dev_trans_start(slave->dev); if ((slave->state == BOND_STATE_ACTIVE) && - (time_after_eq(jiffies, dev_trans_start(slave->dev) + - 2 * delta_in_ticks) || - (time_after_eq(jiffies, slave_last_rx(bond, slave) - + 2 * delta_in_ticks)))) { + (!time_in_range(jiffies, + trans_start - delta_in_ticks, + trans_start + 2 * delta_in_ticks) || + !time_in_range(jiffies, + slave_last_rx(bond, slave) - delta_in_ticks, + slave_last_rx(bond, slave) + 2 * delta_in_ticks))) { + slave->new_link = BOND_LINK_DOWN; commit++; } @@ -2956,6 +2976,7 @@ static void bond_ab_arp_commit(struct bo { struct slave *slave; int i; + unsigned long trans_start; bond_for_each_slave(bond, slave, i) { switch (slave->new_link) { @@ -2963,10 +2984,11 @@ static void bond_ab_arp_commit(struct bo continue; case BOND_LINK_UP: + trans_start = dev_trans_start(slave->dev); if ((!bond->curr_active_slave && - time_before_eq(jiffies, - dev_trans_start(slave->dev) + - delta_in_ticks)) || + time_in_range(jiffies, + trans_start - delta_in_ticks, + trans_start + delta_in_ticks)) || bond->curr_active_slave != slave) { slave->link = BOND_LINK_UP; bond->current_arp_slave = NULL;