From mboxrd@z Thu Jan 1 00:00:00 1970 From: Andrew Morton Subject: Re: [Bugme-new] [Bug 16187] New: Carrier detection failed in dhcpcd when link is up Date: Mon, 21 Jun 2010 14:21:38 -0700 Message-ID: <20100621142138.3416b5d0.akpm@linux-foundation.org> References: <20100615142418.f47e8abd.akpm@linux-foundation.org> <20100618061641.GA7039@lackof.org> <8DD2590731AB5D4C9DBF71A877482A900158F12BEA@orsmsx509.amr.corp.intel.com> Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Cc: Grant Grundler , "netdev@vger.kernel.org" , Kyle McMartin , "bugzilla-daemon@bugzilla.kernel.org" , "bugme-daemon@bugzilla.kernel.org" , "casteyde.christian@free.fr" , YOSHIFUJI Hideaki To: "Allan, Bruce W" Return-path: Received: from smtp1.linux-foundation.org ([140.211.169.13]:48833 "EHLO smtp1.linux-foundation.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1758444Ab0FUVWr (ORCPT ); Mon, 21 Jun 2010 17:22:47 -0400 In-Reply-To: <8DD2590731AB5D4C9DBF71A877482A900158F12BEA@orsmsx509.amr.corp.intel.com> Sender: netdev-owner@vger.kernel.org List-ID: On Fri, 18 Jun 2010 08:04:36 -0700 "Allan, Bruce W" wrote: > On Thursday, June 17, 2010 11:17 PM, Grant Grundler wrote: > > On Tue, Jun 15, 2010 at 02:24:18PM -0700, Andrew Morton wrote: > >> > >> (switched to email. Please respond via emailed reply-to-all, not > >> via the bugzilla web interface). > > > > I've resync to linus' tree (2.6.35-rc3) and reviewed the output of: > > git diff v2.6.34 drivers/net/tulip/ > > > > I don't see anything that would affect how link state > > changes get reported to user space. > > > > I'm not inclined to believe this is a tulip "bug" unless > > core netdev behavior changed and tulip is not longer > > doing the right thing. > > > > hth, > > grant > > I don't believe this is a tulip specific bug - the same thing has been reported against e1000e and bnx2 (IIRC); I have not had the time to investigate further. So it's affecting three drivers. One thing which changed in there recently is : commit b2db756449f63f98049587f7ede4a8e85e0c79b1 : Author: YOSHIFUJI Hideaki : AuthorDate: Sat Mar 20 16:11:12 2010 -0700 : Commit: David S. Miller : CommitDate: Sat Mar 20 16:11:12 2010 -0700 : : ipv6: Reduce timer events for addrconf_verify(). So perhaps someone could test the simple reversion patch, below? I couldn't locate these e1000e and bnx2 bug reports so I couldn't cc the reporters :( I'm seeing several patches on netdev "use netif_carrier_off to prevent tx timeout". Is that related? net/ipv6/addrconf.c | 27 ++++----------------------- 1 file changed, 4 insertions(+), 23 deletions(-) diff -puN net/ipv6/addrconf.c~revert-2 net/ipv6/addrconf.c --- a/net/ipv6/addrconf.c~revert-2 +++ a/net/ipv6/addrconf.c @@ -100,10 +100,6 @@ #define INFINITY_LIFE_TIME 0xFFFFFFFF #define TIME_DELTA(a, b) ((unsigned long)((long)(a) - (long)(b))) -#define ADDRCONF_TIMER_FUZZ_MINUS (HZ > 50 ? HZ/50 : 1) -#define ADDRCONF_TIMER_FUZZ (HZ / 4) -#define ADDRCONF_TIMER_FUZZ_MAX (HZ) - #ifdef CONFIG_SYSCTL static void addrconf_sysctl_register(struct inet6_dev *idev); static void addrconf_sysctl_unregister(struct inet6_dev *idev); @@ -3159,15 +3155,15 @@ int ipv6_chk_home_addr(struct net *net, static void addrconf_verify(unsigned long foo) { - unsigned long now, next, next_sec, next_sched; struct inet6_ifaddr *ifp; struct hlist_node *node; + unsigned long now, next; int i; rcu_read_lock_bh(); spin_lock(&addrconf_verify_lock); now = jiffies; - next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY); + next = now + ADDR_CHECK_FREQUENCY; del_timer(&addr_chk_timer); @@ -3181,8 +3177,7 @@ restart: continue; spin_lock(&ifp->lock); - /* We try to batch several events at once. */ - age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ; + age = (now - ifp->tstamp) / HZ; if (ifp->valid_lft != INFINITY_LIFE_TIME && age >= ifp->valid_lft) { @@ -3252,21 +3247,7 @@ restart: } } - next_sec = round_jiffies_up(next); - next_sched = next; - - /* If rounded timeout is accurate enough, accept it. */ - if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ)) - next_sched = next_sec; - - /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */ - if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX)) - next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX; - - ADBG((KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n", - now, next, next_sec, next_sched)); - - addr_chk_timer.expires = next_sched; + addr_chk_timer.expires = time_before(next, jiffies + HZ) ? jiffies + HZ : next; add_timer(&addr_chk_timer); spin_unlock(&addrconf_verify_lock); rcu_read_unlock_bh(); _